FACULTY OF SCIENCE AND TECHNOLOGY

MASTER'S THESIS
Study program:
Spring semester, 2019
Petroleum Technology/ Natural Gas (Open access)
Engineering
Author:

Awais Ashraf

Faculty supervisor:

Prof. Rune Wiggo Time

Title:

Analysis of flame propagation using an imaging technique BOS (Background Oriented Schlieren)
and PIV (Particle Imaging Velocimetry).

Credits: 30 ECTS
Keywords:
Number of pages: 94
Flame analysis, flame speed, flame
+ supplemental material/other: 4
propagation, Background Oriented Schlieren
(BOS), Particle Imaging Velocimetry (PIV).

Stavanger, 15.06.2019
Abstract

An experimental and simulation work has been conducted to study premixed flame propagation
using imaging technique BOS (background oriented schlieren) and PIV (particle imaging
velocimetry). Flame images obtained from experiment and computation using PIVlab 1.41 were
used for studies.

The BOS (background oriented schlieren) technique is a newly established method for easily and
inexpensive visualizing density gradient variation in compressible flows. The BOS setup design
consists primarily of a background with random dots and a camera targeted on the dots.
Variation of the density between the uninterrupted image and the distorted image resulted in a
dots shift that is used by the optical flow method to generate displacement field.

We present a simple 2D imaging method for measuring this distortion and then show how it can
be used to visualize gas and liquid flows. In addition, several research questions associated to
combustion and fluid mechanics are explored using high speed direct/schlieren imaging, PIV
techniques, etc.

ii
Acknowledgement

I would like to pay my gratitude to Prof. Rune Wiggo Time for his consistent support and
motivation throughout the achievement of this thesis. His remarkable and kind benefaction
through word and perception has been of countless worth.

In addition, I would like to thank Senior Engineer Hermonja A. Rabenjafimanantsoa, Benja, for
providing me opportunity to work in laboratory and handed over keys to work in my desire
schedule.

Finally, I want to acknowledge the support, and love of my family – my parents and siblings.
They all kept me going and this work was not possible without their support.

Author

Awais Ashraf

iii
Contents

Abstract ......................................................................................................................................................... ii
Acknowledgement ....................................................................................................................................... iii
List of figures ............................................................................................................................................... vi
List of tables............................................................................................................................................... viii
Abbreviations ............................................................................................................................................... ix
Chapter 1: Introduction ................................................................................................................................. 1
1.1. Project background ....................................................................................................................... 1
1.2. Motivation ..................................................................................................................................... 1
1.3. Project Goal .................................................................................................................................. 2
1.4. Arrangement of the thesis ............................................................................................................. 2
Chapter 2: Literature Review ........................................................................................................................ 3
2.1. Fuels and the combustion process ................................................................................................. 3
2.1.1. Combustion ............................................................................................................................... 3
2.1.2. Premixed flame ......................................................................................................................... 4
2.1.3. Non-premixed/diffusion flames ................................................................................................ 5
2.1.4. Transition from Laminar to turbulent flames ............................................................................ 6
2.1.5. The buoyant plume.................................................................................................................... 7
2.2. Optical experimental techniques in basic research on combustion ............................................... 9
2.2.1. Background Oriented Schlieren and similar techniques ......................................................... 10
2.2.2. Optical principle of BOS method............................................................................................ 13
2.2.3. The advantages of BOS........................................................................................................... 14
2.2.4. The disadvantages of BOS ...................................................................................................... 14
2.3. What is an Image? ....................................................................................................................... 15
2.3.1. Image Analysis problem?........................................................................................................ 16
2.3.2. Image Storage and Compression............................................................................................. 17
2.3.3. Measurements ......................................................................................................................... 18
2.4. What is noise? ............................................................................................................................. 19
2.4.1. Filter ........................................................................................................................................ 19
2.5. Particle Image Velocimetry (PIV) .............................................................................................. 20

iv
 2.5.1. Implementation and architecture ............................................................................................. 21
2.5.2. Image pre-processing .............................................................................................................. 22
2.5.3. Calibration............................................................................................................................... 23
2.5.4. Computation of the cross-correlation function ....................................................................... 24
2.6. Divergence .................................................................................................................................. 24
Chapter 3: Experimental Setup ................................................................................................................... 25
3.1. Experiment and required resources ............................................................................................. 25
3.1.1. Video capturing ....................................................................................................................... 25
3.1.2. Video split program/ software................................................................................................. 26
3.1.3. Data processing/BOS image processing software................................................................... 27
3.1.4. Results evaluation ................................................................................................................... 28
3.2. Challenges:.................................................................................................................................. 29
Chapter 4: Results & Discussion ................................................................................................................ 30
Chapter 5: Conclusion & Future Work ....................................................................................................... 80
References ................................................................................................................................................... 82
Appendix A ................................................................................................................................................. 86
Appendix B ................................................................................................................................................. 94
Appendix C ............................................................................................................................................... 104
Appendix D ............................................................................................................................................... 106

v
List of figures

Figure 1: Premixed burner schematic illustration [4] ................................................................................... 5

Figure 2: Premixed and Non-premixed flame [5] ......................................................................................... 6
Figure 3: Flame height vs jet velocity, showing flow transition to turbulence [6] ....................................... 7
Figure 4: The buoyant plume development [2] ............................................................................................. 8
Figure 5: Diagram of axisymmetric plume [2] ............................................................................................. 8
Figure 6: Digital image processing and its bionics [39] ............................................................................... 9
Figure 7: Development of measurement techniques [23] ........................................................................... 11
Figure 8: Background Oriented Schlieren .................................................................................................. 12
Figure 9: Actual image area with corresponding magnified view to show pixel intensities [24] ............... 16
Figure 10: Image analysis process steps [24].............................................................................................. 17
Figure 11: Filter processing for de-noising [40] ......................................................................................... 19
Figure 12: PIV principle [28] ...................................................................................................................... 20
Figure 13: Workflow of PIV analysis in PIVlab [28] ................................................................................. 21
Figure 14: Noise reduction process [40] ..................................................................................................... 22
Figure 15: The consequence of pre-processing techniques [28] ................................................................. 23
Figure 16: Experiment flow diagram .......................................................................................................... 25
Figure 17: Figure describes inputs (left side) and output of MATLAB program ....................................... 28
Figure 18: Artificial experiment for 0.01 pixel shift produced by using MATLAB program .................... 30
Figure 19: Artificial experiment for 0.05 pixel shift ................................................................................... 31
Figure 20: Input images of the experiment using white paper .................................................................... 32
Figure 21: PIV analysis images of experiment on white paper .................................................................. 33
Figure 22: Schematic diagram of BOS experiment .................................................................................... 34
Figure 23: PIV analysis images 1 at reference time for 30 fps ................................................................... 35
Figure 24: PIV analysis images 2 at reference time for 30 fps ................................................................... 36
Figure 25: PIV analysis images 3 at reference time for 30 fps ................................................................... 37
Figure 26: PIV analysis images 4 at reference time for 30 fps ................................................................... 38
Figure 27: PIV analysis images after 5 sec for 30 fps................................................................................. 40
Figure 28: PIV analysis images after 10 sec for 30 fps............................................................................... 42
Figure 29: PIV analysis images 1 at reference time for 60 fps ................................................................... 44
Figure 30: PIV analysis images 2 at reference time for 60 fps ................................................................... 45
Figure 31: PIV analysis images 3 at reference time for 60 fps ................................................................... 46
Figure 32: PIV analysis images 4 at reference time for 60 fps ................................................................... 47
Figure 33: PIV analysis images after 5 sec for 60 fps................................................................................. 49
Figure 34: PIV analysis images after 10 sec for 60 fps............................................................................... 51
Figure 35: PIV analysis images 1 at reference time for 240 fps ................................................................. 53
Figure 36: PIV analysis images 2 at reference time for 240 fps ................................................................. 54
Figure 37: PIV analysis images 3 at reference time for 240 fps ................................................................. 55
Figure 38: PIV analysis images 4 at reference time for 240 fps ................................................................. 56
Figure 39: PIV analysis images after 5 sec for 240 fps............................................................................... 58
Figure 40: PIV analysis images after 10 sec for 240 fps............................................................................. 60
Figure 41: Effect of surrounding air and flame plume at stable position .................................................... 61
Figure 42: Effect of rolling shutter on moving propeller [37] .................................................................... 62

vi
Figure 43: Diagram depicting experimental setup using mirror ................................................................. 63
Figure 44: PIV analysis images 1 using mirror at reference time for 30 fps .............................................. 64
Figure 45: PIV analysis images 2 using mirror at reference time for 30 fps .............................................. 65
Figure 46: PIV analysis images 3 using mirror at reference time for 30 fps .............................................. 66
Figure 47: PIV analysis images using mirror after 5 sec for 30 fps ............................................................ 68
Figure 48: PIV analysis images using mirror after 10 sec for 30 fps .......................................................... 70
Figure 49: PIV analysis images 1 using mirror at reference time for 60 fps .............................................. 71
Figure 50: PIV analysis images 2 using mirror at reference time for 60 fps .............................................. 72
Figure 51: PIV analysis images 3 using mirror at reference time for 60 fps .............................................. 73
Figure 52: PIV analysis images 4 using mirror at reference time for 60 fps .............................................. 74
Figure 53: PIV analysis images using mirror after 5 sec for 60 fps ............................................................ 76
Figure 54: PIV analysis images using mirror after 10 sec for 60 fps .......................................................... 78
Figure 55: Internal reflection in mirror [38] ............................................................................................... 79

vii
List of tables

Table 1: Types and examples of combustion [3] .......................................................................................... 4

Table 2: Comparison between number of dots and diameter for background pattern ................................ 26

viii
Abbreviations

BMP Bitmap
BOS Background Oriented Schlieren
CCD Charged Coupled Device
CFD Computational Fluid Dynamic
CLAHE Contrast Limited Adaptive Histogram Equalization
CMOS Complementary Metal-Oxide Semiconductor
DCC Direct Cross Correlation
FFT Fast-Fourier Transform
FPS Frame per Second
GUI Graphical User Interface
ICE Internal Combustion Engine
JPEG Joint Photography Experts Group
PDFA Phase Detection Autofocus
PIV Particle Imaging Velocimetry
PLIF Planar Laser Induced Fluorescence
PNG Portable Network Graphics
Re Reynolds number
TIFF Tagged Image File Format

ix
Chapter 1: Introduction

1.1. Project background

Since ancient times, human beings are using flame and fire to fulfill the different daily purposes.
Today this process of burning is still being extensively used in industrial applications.
Approximately 90% of the usage of this energy is linked to combustion like processes, for
instance gas turbine engine and ICE (internal combustion engine) etc. Researchers have
developed many models and methods to comprehend the behavioral mechanism of this
complicated phenomenon. Combustion plays an important role in modern industry.
Understanding of the combustion process is a quickly emerging area and it involves a broad
variety of subjects like flame dynamics, ignition, flame quenching and the mechanism of
chemical reaction.

In experimental fluid dynamics, generally three approaches are being used to visualize the flow:
visualization of surface flow, particle tracer techniques, and optical methods. Surface flow
visualization discloses the flow streamlines as the solid surface methodologies. One example of
this technique is colored oil applied to the surface of a wind tunnel model. Particles like smoke
can be added to a flow to trace flow movement. A part of the complicated flow pattern could be
illuminated by laser sheet to understand this complex phenomenon. By Assuming that the
particles track the streamlines of the flow closely, we can not only visualize the flow but also
measure its velocity using a method known as PIV (particle image velocimetry). Finally, by
changing their optical refractive index, some flows reveal their patterns. These techniques help to
visualize the phenomenon using optical techniques generally known as the schlieren,
shadowgraph, interferometry and more recently, the BOS (background oriented schlieren)
technique.

1.2. Motivation

Optical diagnosis plays an important role among all combustion diagnostic techniques as it
appears to have certain advantages like it is intuitive, non-intrusive and offering large
information sets contrast to single point measurement. In recent times, for both quantitatively
and qualitatively visualization of combustion processes, optical techniques such as

1
shadowgraph/schlieren, PIV and PLIF, have been extensively used. The understanding of
combustion characteristics can be increased by combining optical visualization, high-speed
imaging and image processing. Most optical techniques are based on 2D (two-dimensional) and
it is fairly new concept to measure flame dynamics in 3D (three-dimension) space.

The objective of this thesis is to develop the BOS visualization method for flame analysis using a
simple optical setup. Another objective of this study is to achieve natural understandings into the
phenomena of combustion/fluids flows using a mixture of optical approaches.

1.3. Project Goal

The purpose of this project was to develop a background oriented schlieren technique for flame
propagation analysis. A BOS method will allow not only capturing qualitative images of
compressible flow features but also extracting quantitative information about the flow.

1.4. Arrangement of the thesis

This thesis has been organized according to guideline provided by the University of
Stavanger. The dissertation has been broke up into five chapters. Chapter 1 highlights goal and
motivation. Chapter 2 presents theories related to Combustion, flame, BOS (background oriented
schlieren), Image, Image Analysis and PIV (particle imaging velocimetry). Chapter 3 represents
experimental setup. In chapter 4, I have represented PIV results and discussion. In the end,
chapter 5 presents conclusion of the thesis.

2
Chapter 2: Literature Review

Combustion involves chemical reaction of air (oxygen) and any combustible material (solid,
liquid or gas), which results fire. It offers heat and power for our domestic and industrial use, but
it could have serious effects on the human health and environment. In 1993, United Kingdom
suffers around 800 human lives and around £1000 million [1]. But over the past two decades,
there is no significant increase in losses due to enhancing technical competence of fire service
and smart fire defense systems.

2.1. Fuels and the combustion process

Fuel can be anything which is burning regardless of state of matter like methane, petrol or a
wooden item in room. Under proper condition, all will burn producing combustion products and
discharging heat.

2.1.1. Combustion

“Combustion is a complex sequence of exothermic chemical reaction between a fuel and an

oxidant, during which heat and light are emitted to form a flame”. A flame, which is seen as a
visible portion, is a self-reliant circulation of a localized combustion zone. In a typical chemical
reaction, for example:

Fuel + Oxidizer → Products of combustion + Energy Equation no. 1

A comprehensive overview of combustion types, flow types and their examples has been
summarized in Table 1. From the Table 1, it can be seen that combustion system can be
classified on the basis of preparation of the fuel (or oxidizer mixture). There are two major types
of mixing; premixed and non-premixed. Their flow could be laminar or turbulent. Bunsen and
flat flame are examples of laminar flow, while gas turbine and spark ignition are examples of
turbulent flow [3].

3
 Table 1: Types and examples of combustion [3]

Type of mixing Flow type Example

- Bunsen flame
Laminar
- Flat flame
Premixed
- Gas turbine
Turbulent
- Spark-ignited engine
- Wood fire
Laminar
- Candle
Non-premixed
- Diesel engine
Turbulent
- Pulverized coal combustion

2.1.2. Premixed flame

Premixed flame is a combustion reaction that occurs before chemistry starts when fuel and
oxidizer are mixed. The flame front movement is a fluid convection and flame propagation
superposition. A simplified example of premixed flame is a Bunsen burner is illustrated in
Figure 1. Chemical process dominates the flame speed of premixed flame for laminar flow of
reactants. Hence, a stationary flat flame front will form, if the flowrate of reactant is equal to the
flame speed. If the flow rate exceeds the velocity of flame, a conical shape will be displayed in
the flame. If the flow rate is lower than the speed of flame, the front moves upstream until the
blend exceeds its flammability boundaries or the flame is quenched owing to flow conditions
modifications. One of main hazard with premixed flame is flame flashback which eventually
leads to explosion [4]. As the speed of flame propagation increases, flame tends to move from
laminar to turbulent flame which increases mass and heat transfer.

4
 Figure 1: Premixed burner schematic illustration [4]

2.1.3. Non-premixed/diffusion flames

Diffusion flame’s main characteristic is that fuel and oxidizer are originally distinct and
combustion happens in the mixing area. A candle or Bunsen flame with the air inlet port closed
is an instance of diffusion flame. The appearance of the flame will hinge on fuel’s nature and the
speed of fuel jet in relation to the surrounding atmosphere [4].

The blending and combustion of fuel and oxidizer takes place simultaneously in non-premixed
flames. At the fuel and air interface, the diffusion flame front is formed. Diffusion flames are
generally mixed-controlled or diffusion-controlled, rather than in premixed flames kinetically-
controlled [3].

5
 Figure 2: Premixed and Non-premixed flame [5]

2.1.4. Transition from Laminar to turbulent flames

Combustion could in occur in laminar or a turbulent flow environment, but it can also take place
in laminar-to-turbulent transitional flow condition. Laminar flow occurs when fluid flows
without lateral mixing in parallel layers between the layers. But turbulent flow is categorized as
chaotic, where adjacent layers constantly upsetting each other. Viscous force dominates the
laminar flow (low Re), while inertial force controlled the turbulent flow (high Re). The flow is
partially laminar and partially turbulent in some circumstances and transition takes from laminar
to turbulent flow in this condition [7].

The behavior of the flame could be influenced by two parameters, flow conditions and buoyance
forces induced by chemical reaction process. With the increase in Re, it is possible to observe a
shift from laminar to turbulent flow. The flow completely changes to turbulent with the increase
in Reynolds number. With the more increase of Re, a flame which is detached from nozzle can
be established. At a critically high value of Re, the flame can be finally blown-off [7].

6
 Figure 3: Flame height vs jet velocity, showing flow transition to turbulence [6]

When jet velocity increases, turbulence starts from top of flame and extends to bottom nozzle but
will never touch it (figure 3). In addition, the influence of air rises through eddy mixing, which
decreases flame height from laminar section to constant value in fully turbulent system.

2.1.5. The buoyant plume

The notion of buoyancy always presented in association to natural convection. If there is a

distinction in densities between liquids, then the buoyancy force will raise the fluid with less
density compared to its atmosphere. To define this convective column, rise from a source of heat,
the term ‘buoyant plume’ is used. Plume contour is defined by its interaction with the
neighboring fluid (mostly air) (figure 4) [7].

7
 Figure 4: The buoyant plume development [2]

The ideal plume is axisymmetric, extending vertically to a level where force of buoyancy
becomes too feeble to overcome viscous drag (figure 5). A temperature inversion can form under
specific atmospheric environments that will cause rising smoke plume to spread laterally (i.e.
interaction with ceiling) [2].

In the flame plume, the vertical movement of buoyant gasses causes entraining of the
surrounding air. This not only provides oxygen/air for fuel to be combusted, but it also cools and
dilutes fire products when they grow above the flame and cause a constant rise in the size of
smoke produced by the fire.

Figure 5: Diagram of axisymmetric plume [2]

8
2.2. Optical experimental techniques in basic research on combustion

Leonardo da Vinci (1452-1519) identified observational science in experiments during

Renaissance era. He was the first person to explain flow visualization phenomena, such as eddies
formation and wave propagation, etc. [7]. Different elements are launched into the fields as a
tracer or contrast medium in traditional flow visualization techniques to discriminate the object
to be observed. To achieve high contrast and resolution in images, flow field needs to be
illuminated. Recently, the flow visualization approaches have advanced to a new era with the
speedy progress of computational methods and computer color graphics. Figure 6 illustrates the
bionics and digital image processing procedure.

(a) Bionics image processing by eye and brain

(b) Digital image processing

Figure 6: Digital image processing and its bionics [39]

9
Figure 6, (a) represents the identification of an object with human eye. In order to focus on the
retina, light rays reflected from the object while passing through the lens. Light rays in the retina
are stimulating millions of optical receptors. Figure 6, (b) shows the digitalization of image and
digital image processing process. For an object to be automatically identified, a system similar to
human vision is needed. A camera is analogous to the human eye and captures the image of
object. A coding system converts the camera image into digital image, a microprocessor
performs the processing function. The output could be softcopy or a hardcopy (printed version)
[7].

The advancement in the optical investigative techniques has played crucial role in flow
visualization, pressure and distribution of temperature, velocity measurements, etc. The results of
visualization help to understand physics of the complex phenomena of flows, encourage theory
articulation and also offer significant feedback for engineering designs.

Optical diagnostic techniques have been applied in combustion research from decades. The
invisible flow pattern significantly affects the ignition and combustion in combustion phenomena
in addition to visible light emission. Combustion includes chemical reactions and excessive
temperature, non-intrusive optical diagnostics therefore demonstrate the huge benefits of
obtaining visualization and measurement. One of the influential flow visualization tools is
schlieren and shadowgraph technique that can visualize invisible variations in density and has
been extensively used to study combustion, fluid mechanics and aerodynamics, etc. Particle
imaging velocimetry is used by introducing tiny particle into flow field as a standard technique
for measuring velocity. PLIF (planer laser induced fluorescence) can tackle specific species
distribution by recognizing the appropriate light emission. Most diagnoses are two-dimensional.
In diagnostic techniques such as stereo imaging and tomography, great efforts have been made to
obtain information in three dimensions. In exploring and developing three dimensional
diagnostics, great efforts are needed [7].

2.2.1. Background Oriented Schlieren and similar techniques

The BOS technique is comparable to observing the effect of local density gradients like heat
haze, mirage or fata morgana, where the observation of the observer distorts a distant object or

10
even mirror the perceived image Figure 7 (a). Another effect, known as shadow effect (Figure 7
(b), is observable with the naked eye but with optical instrumental aid it became more evident.
The density gradient between the source of light and the surface with monotonous brightness and
color generates this effect. The changes in intensity ∆I at every point on the surface are directly
proportional to the second air density derivative. The shadowgraph is known as the picture of
theses varying light intensities [23].

Techniques for measuring optical density such as shadowgraphy, schlieren or interferometry are
well established and have been in use for many centuries. They are based primarily on variation
in the refractive index and allow flow density variations to be observed [8].

(a) Mirage or heat haze (b) Shadowgraphy (since the 17th century)

(c) Toepler Schlieren photography (since the (d) Laser speckle photography (since the 20th
19th century) century)

Figure 7: Development of measurement techniques [23]

11
The study of the density gradients became important in the field of fluid dynamics when
aerodynamics developed transonic. An important driver for enhancement and application of
observation of fluid dynamics was the experiment of bullets shooting and the study of flow fields
in the work of Ernst Mach. This method is most frequently used for the visualization of fluid
density in Toepler’s schlieren photographical work (Figure 7, (c)). Typically, its experimental
configuration comprises of spherical mirrors and an aperture. The light intensity variation ∆I
captured here are proportional to the first fluid density derivative.

Laser speckle interferometry is another method that is very close to BOS method (Figure 7, (d)).
The screen was illuminated with laser light by the density gradient under investigation, which
produces a speckled interference pattern. The generated and optically evaluated displacement ∆s
is proportional to the first density derivative. The schlieren technique is typically centered on the
valuation of image displacements. Incoherent light can illuminate the background and imaged it
through a fluid. Illustration of BOS is presented in Figure 8.

Figure 8: Background Oriented Schlieren

The experimental arrangement for typical arrangements of the BOS experiment can be organized
by putting the background target with the randomly generated dots on one side and a camera on
the opposite side [10, 11]. The background target is important in image capturing and can affect
the spatial resolution [11, 12]. Many factors can influence the measurement and most important
factors are the camera’s aperture, focal length, distances between the optical elements and the
size of the interrogation window [11, 13, 15, and 16].

12
Digital valuation and white-light illumination are also well-known in speckle interferometry, So,
BOS could also be known as “white-light speckle density photography”. However, the most
common name for schlieren method is “background oriented” that actually explains how the
camera focuses on objects behind the investigated flow [9].

2.2.2. Optical principle of BOS method

In the present research, the basic idea of this optical technique is based on the variation of air
refractive index due to density gradients. The Gladstone-Dale equation explains the relationship
between density variations and refractive index:

𝑛−1
= 𝐺(𝜆) Equation no. 2
𝜌

Where n: refractive index, ρ: density (kg/m3) and G (λ): Gladstone-Dale number (m3/kg).

The method is best compared to density speckle photography as explained by Debrus et al.
(1972), Kopf (1972) and Wernekinck and Merzkirch (1987) in an updates edition. The expanded
laser beam and a ground glass generates a pattern, a laser spackle pattern is replaced by a
background pattern with random dots. The laser pattern should have the high spatial frequency
that could be helpful to have images with reasonable distinction.

In the first step, a reference image is generated by recording the background pattern detected
through air at rest. Additional exposure through the flow under analysis results in locally
displaced background pattern image in the second step. Then both resulting images by the
exposures can examine by methods of image correlation. Existing assessment algorithms, for
instance for particle image velocimetry, which have been established and optimized, can then be
used to determine pattern displacement at various positions.

The main distinction between BOS method and the other techniques described above is that no
optical devices are needed to produce background. The only optical tool required is a lens on the
camera. The camera’s main focus is the background pattern that produces an image that is
moderately analogous to a particle image and that’s why the technique is known as “background
oriented” [8].

13
In different studies, similar methods of observing and comparing background pattern were
observed [12, 17]. If the spatial resolution of the background does not matter, the only main
requirement for the target is a high pattern contrast [12] that can be readily accomplished by
printing dots on a paper with a standard printer. The literature also recorded the effective use of
natural background patterns [10, 11, and 18]. BOS is able to measure two in-plane, first-order
gradients parallel to the background target and the image sensor, analogous to the schlieren
method, which are integrated along the line of sight [13, 14, 16, and 19].

2.2.3. The advantages of BOS

It not only provides the quantitative measurement, but it also has some other benefits and can be
summarized as: simpler modification compared to traditional schlieren methods, reduced number
of optical elements, no restriction on measurement object size and straightforward computation
[11, 13, and 14]. The potential of using natural backgrounds to image is an advantage of BOS
[18, 21]. In addition to temperature or density measurement, the mixture of PIV and BOS can
provide valuable data regarding velocity circulation in a flow.

2.2.4. The disadvantages of BOS

(a) Limited resolution:

BOS always delivers less image resolution than comparable traditional schlieren methods
because image cross-correlation algorithms average their interrogation window size [23].
Similarly, an increase in BOS sensitivity requires a larger interrogation window, resulting lower
resolution [13]. In discrete flow features such as shock waves, resolution loss is most noticeable.

(b) Difficulty in attaining sharp focus on both subject and background:

It is clear that camera sensor needs to have a sharp focus on the BOS background [23], but what
about schlieren object (in my case flame) under study? The camera’s depth-of-field can
theoretically extend to include both, but this can lead to an impractically large distance from
camera to schlieren object, especially in a laboratory where dimensions are restricted. In order to

14
minimize this problem, Meier [11] recommends a small camera aperture and a short distance to
the background from the schlieren object, although the latter also reduces BOS sensitivity.

(c) Non-parallel illumination:

BOS shares the problem of having a small camera lens looking at a large background grid along
non-parallel rays with direct shadowgraphy and lens-and-grid schlieren. This complicates the
quantitative measurement of schlieren object. Avoiding these pitfalls is an important reason for
expensive lenses and mirrors in schlieren instruments.

(d) Vibration sensitivity:

BOS compares a flow-on image to a previously acquired on-flow image. Vibration, or anything
other than schlieren object that changes between these two exposures, is a problem [13].

BOS observations are extremely sensitive in the experimental instruments to any changes and
deflection. Any disturbance or vibration on the optical path may result in elevated evident
background pattern displacement and trigger BOS signal deformation. Precautions should
therefore be taken to eradicate any vibration in the optical path.

2.3. What is an Image?

An image is a three-dimensional array of numbers in its simplest form that shows intensities and
spatial coordinates (x and y- axis) of pictured item (figure 9). To improve or make numerical
measurement of attributes in an image, mathematical calculations are carried out on numbers of
an array. In the digital world, pixels are typically small square picture element, by which an
image is composed. The number of light photons that strike a camera’s detector determines each
pixel’s intensity or gray level. Typically, images ranges from 256*256 pixels to 4096*4096
pixels using specialized imaging accessories. Resolutions and sensitivities of cameras are getting
better day by day.

15
 Figure 9: Actual image area with corresponding magnified view to show pixel intensities [24]

2.3.1. Image Analysis problem?

The process of analyzing the visualized scene incorporates specific steps that lead to an
improved image or data that can be used for further interpretation. A conclusion is essential at
each step to be able to accomplish the next step, as shown in figure 10. Moreover, different
algorithms can be used for every step to accomplish the anticipated effects and/or measurements.

16
 Figure 10: Image analysis process steps [24]

2.3.2. Image Storage and Compression

Many systems save visual data onto a permanent medium (e.g., floppy, hard, and optical disks)
using exclusive algorithms that compress the images to some extent. There are standardized
compression algorithms, like, JPEG (joint photography experts group) and TIFF (tagged image
file format). These algorithms make it heavy to transport the images, but several systems allow
exporting images into standard formats. When the images are stored in standard formats,
considerable loss of information could occur because of compression of images, so, great care
should be exercised when storing images.

Therefore, it is generally recommended that the images for analyzing purposes should be stored
with very limited compression (no less than half the original size) or no compression as
compared to the images that are for archival and visualization purposes only that could be
compressed as desired. The information loss connected with compression of images depends on

17
the type and size of image being compressed and gray-level range of the structures to be
quantified.

There are four commonly used image formats, JPEG, TIFF, BMP, and PNG. JPEG is
commonly used in domestic devices. Images can be very large these days with good quality
cameras around. When image is compressed to different formats, following is the sequence
according to size:

BMP > Tiff > PNG > JPEG

 JPEG achieves very good compression ratios, but is lossy. Take an image, put it through
JPEG compression process, you'll get a very small image, but it will be impossible to
recover from that image the one that you started with.
 TIFF, the tanked image file format, is generally, though not always, lossless. TIFF has
many settings and options, and it's possible to create a lossy version of TIFF very, very
easily. So TIFF should be used with care.
 PNG is a lossless format. So there is no danger with a PNG images that you will destroy
an image in the compression process.
 BMP is perhaps the simplest of the available formats. It's lossless so it's safe to use, but
doesn't allow you to add any metadata or any layers. In most scientific applications, BMP
or PNG are the safest to use.

2.3.3. Measurements

The essence of image analysis is making a measurement or series of measurements that quantify
some aspect of the image. Image analysis measurements can be made manually, automatically,
or even by comparing a series of images that already have been measured. Manual measurement
usually involves counting features, such as points, intersections, and intercepts, but it also
involves measuring length. Area and volume, on the other hand, usually are derived values. All
of these measurements are easily calculated by the computer for every feature in the image(s).

18
2.4. What is noise?

Whenever we're taking images for use in scientific experiments, we're faced with a number of
problems. One of these is the camera or the microscope or whatever device it is that we're using
to take the images doesn't actually capture exactly what's out there in the real world. Whatever
we're seeing in the image has been distorted ever so slightly by the way that we're capturing it
and the device that we're using and the process by which that device turns the lights on the signal
into some kind of digital data. This destructive process can be called image noise, where we're
distorting that signal that we're recording [40].

To measure and estimation of the noise level appears to be an impossible task from a single
image: it should be acknowledged that whether texture, color or lighting from image itself is the
cause of the image variations, or whether it is due to noise. Real CCD camera noise,
unfortunately, depends heavily on the level of image intensity [27].

2.4.1. Filter

Noise reduction processes usually take the form of filters. These work by examining a small
window of pixels called a kernel, for example 3 x 3 pixels in size, then recalculation the middle
pixel based on the neighboring pixels in this window. This window is then moved across the
image typically a pixel at a time and applied again until the whole image has been recalculated.
We can perform different calculations on the pixels in the kernel to handle different kinds of
noise, but in all cases, we are using local image statistics (such as a mean) to recalculate the
center value based on the context of its neighbor’s [40].

Figure 11: Filter processing for de-noising [40]

19
2.5. Particle Image Velocimetry (PIV)

PIV (Particle image velocimetry) is a widely accepted measurement technique for studying
turbulent flows structure since it offers to measure quantitative data on the velocity field’s spatial
structure.

Particle image velocimetry (PIV) is an accepted approach for visualization of non-intrusive,

qualitative and quantitative flow. Here, for PIV analysis in MATLAB, a GUI-based open-source
tool (PIVlab) was used. Multiple embedded MATLAB features are used and subsequent
computing is facilitated by providing a close connection to the common MATLAB [28].

An illustration of PIV principle has been shown in Figure 12. In PIV, fluid movement (either
liquid or gaseous) is observed by enlightening a small sheet of fluid having neutrally buoyant
and reflective footprint particles. Parallel sheet position is given for digital image sensors to
capturing particle movements. Two images (A and B) of the sheet are captured at to and to+∆t for
PIV analysis. From distance and difference in time (∆t) between two images, velocities can be
derived from the illuminating sheet. The particle displacement is determined by assessing the
cross-correlation of many small sub-images (interrogation areas). The correlation provides the
most likely particle shift between image A and image B moving on a straight line [29].

Figure 12: PIV principle [28]

20
2.5.1. Implementation and architecture

MATLAB has a lot of program applications and PIVlab is one of the image processing
applications. PIV analysis involves three main steps (pre-processing, image evaluation and post-
processing, figure 13).

•Images
Input

•CLAHE
•Intensity capping
Pre-processing •Intensity high-pass

•Time
Calibration •Distance

•DCC
•DFT
Image evaluation •Peak finding

•Validation
•Interpolation
Post-processing •Smoothing

•Extraction
Data exploration •Derivatives

•Images
Export •Vector maps

Figure 13: Workflow of PIV analysis in PIVlab [28]

GUI based interface of PIVlab has all of these steps. Workflow is menu-based, starts with the
image input and continues to the bottom of the figure.

21
 An overview of relevant technique and features will be provided in the following section:

2.5.2. Image pre-processing

A general way for better measurement is to improve the images before the actual image
correlations used [29, 30]. Some of the pre-processing techniques are given below: (figure 15)

(a) Histogram equalization:

To enhance the readability of the image data in medical imaging, Contrast limited adaptive
histogram equalization (CLAHE) was established [31]. CLAHE works on small sections of the
image (sections are called tiles or window). In each window, the image histogram intensities
with highest frequencies are extended to the full data spectrum (in 8-bit images from 0 to 255)
(figure 14). High exposure (high intensity values) and low exposure sections are optimized
independently. CLAHE ameliorate the likelihood of identifying valid vectors by 4.7 ± 3.2% [30].

Figure 14: Noise reduction process [40]

(b) Intensity high-pass:

High-pass filter maintain the high-frequency information from the particle enlightenment, mostly
used for inhomogeneous lighting which cause low-frequency background information. The filter
accentuates the image element data and burkes all image low-frequency data (including all
information on low-frequency movement) [28].

22
 (c) Intensity capping:

The PIV technique presumes the same motion for all particles in the examination window. As
perfectly uniform flow probably not exists, this will not be the case in reality. Bright points or
dots in the window numerically share more to correlation signal that result in non-uniform flow
[30]. This problem is avoided by intensity capping filter. The upper boundary of the intensity of
grayscale is selected, and this upper limit replaced all pixels exceeding the threshold. Preventing
the possible negative effect of image adjustments, only a slight quantity of the pixel intensity
information is attuned in intensity capping unlike CLAHE [30]. The prospect of finding valid
vector is enhanced by 5.2 ± 2.5% by intensity capping [30].

Figure 15: The consequence of pre-processing techniques [28]

2.5.3. Calibration

The image analysis in PIVlab calibration was required. Without calibrating the analysis, software
will measure displacement vectors in pixels/second units. To calibrating an image, two
measurements are required; Length of background and time between frames (that can be
calculated using frames per second speed, also shown calculation in chapter 4).

For 30 fps video, time is calculated in the following way:

1
𝑡=
30

𝑡 = 33.3 ∗ 10−3 sec = 33.3 𝑚𝑠𝑒𝑐

23
2.5.4. Computation of the cross-correlation function

Most commercially available data processing software is based on cross-correlation function

Fast-Fourier Transform (FFT) algorithms. Most of these software are iterative and primarily
grounded on an initial evaluation of velocity vectors. They use large interrogation windows
related to a high SNR owing to the huge quantity of particles taken into account in the statistics,
but compromised with poor spatial resolution. The dimensions of the interrogation window is
also gradually reduced in practice to reach a final size of 32x32 or 16x16 pixels that depends on
the density of particle images.

Typically raw images undergo pre-processing before the application of cross-correlation

computation to improve contrast, decrease background noise (e.g. generated by particles deposits
on windows), or to standardize the intensities of the particle to evade the bias of the cross-
correlation function towards bulky particles.

2.6. Divergence

Divergence is the quantity of flux inflowing or exiting a point or the quantity diverging from a
given volume. This could be defined as the rate of flux enlargement (positive divergence) or flux
shrinkage (negative divergence). For example, If A is velocity of the fluid, if ∇•A (divergence of
vector field A) is positive at a point, it means the expansion of fluid at a point or density at that
point is falling with time. If ∇•A is negative, it implies that either the fluid is shrinking and its
density is increasing at the point or the point is the sink of the fluid [41].

𝐹𝑙𝑢𝑥
𝐷𝑖𝑣𝑒𝑟𝑔𝑒𝑛𝑐𝑒 = Equation no. 3
𝑉𝑜𝑙𝑢𝑚𝑒

Divergence of a vector quantity is a scalar [42].

 Positive divergence value implies the expansion of vector.

 Negative divergence value suggests the vector is converging.

24
Chapter 3: Experimental Setup

3.1. Experiment and required resources

In this chapter, an experimental set up was made with necessary instrumentation to evaluate the
flame propagation using BOS (background oriented schlieren) and PIV (particle imaging
velocimetry). Figure 16 presents a comprehensive overview of experimental setup; details of
each step have been described below.

Video capturing

Result Split into

evaluation images

Image processing
using software
(PIVlab)

Figure 16: Experiment flow diagram

3.1.1. Video capturing

A smartphone Samsung Galaxy S8 SM-G950F was used to capture the videos for
experiment. This smartphone has latest Android 9 operating system (OS). This smartphone is not
the latest available model, but one of the recent smartphones with high resolution camera
technology. This smartphone has main camera of 12-megapixel Dual pixel PDAF (Phase
detection autofocus) with aperture of f/1.7. It can record videos at 3 different resolutions with
different frames per second (2160p@30fps, 1080p@60fps, 720p@240fps).
25
a) Accessories:
 A premixed gas (propane-butane) burner was used.
 Two spot light were used to illuminate the background.
 Tripod for stability.

b) BOS background program:

Background is a random dot pattern image generated by a computer and printed on an A3

or A4. The pattern image was produced using the “makebospattern.m” file, which is integrated in
PIVMat Toolbox (i.e. also included in appendix A) [34]. In Table 2, a comparison of number of
dots generated and diameter for background pattern has been presented.

Table 2: Comparison between number of dots and diameter for background pattern

No. of Dots ‘n’ Diameter ‘d (mm)’ Remarks

50000 0.25 Very low visibility
75000 0.25 Very low visibility
100000 0.25 low visibility
125000 0.25 low visibility
150000 0.25 Good visibility
50000 0.5 low visibility
75000 0.5 low visibility
100000 0.5 Good visibility
125000 0.5 Good visibility
Very Good visibility (Best choice
150000 0.5
for analysis)

From the above comparison, it could be concluded that best option for analysis is to used
n = 150000 and d = 0.5 mm, to visualize the displacement of background.

3.1.2. Video split program/ software

Each animation or video consists of a series of still images. Then, several times a second,
these images are played one by one, which fools the eyes into thinking that the object moves.
The faster the images are played the more fluid and smoother the movement appears.

26
 Each image is referred to as a frame that shows the term fps (frames per second). A video
file on a computer simply stores all the frames together and plays them in order, and hundreds of
thousands of the total frames stored for a typical movie. It’s pretty easy to capture an image of
one or two frames, and you have to pause the video and press the Print Screen key. It is
incredibly inefficient and time consuming to capture images one by one, if you want to extract a
range or succession of frames or even all frames from a short video. For that purpose, a program
is needed that can extract frames from a video and automatically save them as image files, such
as jpg, png or bmp. There are a lot of software’s available for video split into images (like VLC
media player, VirtualDub etc.) and also this can be done on MATLAB script program. I have
used MATLAB script program for video splitting (appendix D).

3.1.3. Data processing/BOS image processing software

After receiving the image pairs to a computer, BOS processing was done and then
processed through PIV image-correlation software. There are a lot of different software’s
developed for image processing. One of them is ImageJ which is developed by the US NIH
(National Institutes of Health) [36], but a smarter GUI (Graphical User Interface) based BOS and
PIV image analysis program is PIVlab, which has been used for current work [28].

Beside PIV analysis, a MATLAB script program was written to get absolute difference
between distorted and reference without disruption images (appendix C). From this analysis, we
can get flame structure (figure 17).

27
 (a) Reference image (without
disturbance)

(b) Distorted image (c) Resulted image by absolute difference

Figure 17: Figure describes inputs (left side) and output of MATLAB program

3.1.4. Results evaluation

Results have been discussed and evaluated in chapter 4.

28
3.2. Challenges:

Below is the list of few challenges faced during the experimental setup and experiment.

 Proper light distribution on background.

 Program to make random number of dot for background.
 Optimum background with sufficient number of dots and diameter to see displacement of
background.
 Slow speed video camera (more frames per seconds) to gather more information from
images.
 Calibration of images to get real unit data from the PIVlab analysis.
 Colors scale/legends on analyses images; color scale shows measure/degree of shift pixel
between two exposures.

29
Chapter 4: Results & Discussion

This chapter describes and presents the analysis of experiments followed by a discussion on the
research results. The results relate to the research questions that led the study. Data were used
investigated to describe flame propagation using background oriented schlieren technique.

4.1. Artificial Experiment:

An artificial experiment was performed to check ability of software (PIVlab) to compute

displacement vectors from input images. In this experiment, MATLAB program was used to
produce images of dot background (black color) and seeding particles (red color). MATLAB
script for seeding particles with dot pattern has been presented in Appendix B. Background dots
were kept fixed and seeding particles were generated in the same way using rand function.

4.1.1. 0.01 pixel shift:

Figure 18 shows the raw images for 0.01 pixel shift produced by using MATLAB
program.

(c) PIV analysis of 0.01 shifts

(a) raw image 1 (b) raw image 2
produced by PIVlab software

Figure 18: Artificial experiment for 0.01 pixel shift produced by using MATLAB program

30
As shown in Figure 18, (a) seeding particles in image have some randomly generated certain
position (xi, yi; i = 1 …. N), and then all of particles shifted to a certain amount (0.01 pixel shift)
of pixels along y-axis (in the same direction, upward) in Figure 18, (b). For that experiment, two
images with same background and position of the seeding particles were experiencing 0.01
(figure 18) and 0.05 (figure 19) pixels shifts along y-axis.

4.1.2. 0.05 pixel shift:

(c) PIV analysis image of 0.05

(a) 0.05 pixel raw image 1 (b) 0.05 pixel raw image 2
pixel shift

Figure 19: Artificial experiment for 0.05 pixel shift

In raw images BOS particles (black background dots) are still at the same place while
seeding particles (red dots) shifted along y-axis, 0.01 and 0.05 pixels respectively. In figure 18,
(c) and 19, (c), all the seeding particles move along y-axis but in PIV analysis, it is apparent that
some vectors are displaced in different directions. This disarray is due to the collision of BOS
and seeding dots and it could be concluded that PIV system correlation has problems in
determining what exactly the movement is.

In a way, the result was excepted because BOS pattern is static and has zero
displacement. When all the seeding particles which are moving upwards because of interaction
with background dots/pattern (BOS particles) look like they move sideways or downwards. They

31
neglect real movement and showing high magnitude movement. From figure 18 and 19, we can
analyze that when the pixel shift is small (0.01 pixel shift), more vectors are scattered in different
directions as compared to 0.05 pixel shift.

4.2. Experiment without background pattern:

First, experiment using white paper was done i.e. without background pattern to check
applicability of BOS process and to see the possibility of application of BOS method on white
background. Figure 20 portrays the input images to PIVlab (minimum two images to get
displacement vectors) obtained using experiment.

a b c

Figure 20: Input images of the experiment using white paper

32
 (i) (ii)

Figure 21: PIV analysis images of experiment on white paper

For one PIV analysis image, we need two input/raw images. Some important results that
could be drawn from the figure 21 are:

 Flame structure movements cannot be correlated because there’s no background pattern

behind the flame and no change in reflective index as BOS is based on reflective index.
 One of the advantages of background pattern is that we can correlate the structure using
relative movement. If strong light will be used on object and looked on background
(white paper), then there’s possibility to get the shadow of the object. This process is very
close to shadowgraphy which is out of scope of this study.

4.3. Experiment on BOS background:

An experiment was done for the study of flame propagation using background oriented
schlieren technique (Figure 22). Images from the experiment then analyzed by particle imaging
velocimetry using PIVlab software to obtain different parameters (like velocity magnitude,
vorticity etc.) and structure of the flame. This experiment was performed by varying frames per
second speed on camera (30 fps, 60 fps and 240 fps used).

33
 Figure 22: Schematic diagram of BOS experiment

4.3.1. 30 fps:

For 30 fps video, calibration was done using following time and length that can be
measured from background:

1
𝑡=
30

𝑡 = 33.3 ∗ 10−3 sec = 33.3 𝑚𝑠𝑒𝑐

Start ignition t = 0 (taken as reference for time):

Following are figures shown in sequence for every experiment, first at t = 0 then after 5 and 10
sec of ignition to portray development of flame structure and upward flow from start to fully
developed at 30 fps.

34
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 23: PIV analysis images 1 at reference time for 30 fps

35
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 24: PIV analysis images 2 at reference time for 30 fps

36
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 25: PIV analysis images 3 at reference time for 30 fps

37
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 26: PIV analysis images 4 at reference time for 30 fps

38
After 5 sec from reference time:

(a) Velocity magnitude (A1) Divergence (A2) Vorticity (A3) U-component (A4) V-component

(b) Velocity magnitude (B1) Divergence (B2) Vorticity (B3) U-component (B4) V-component

39
(c) Velocity magnitude (C1) Divergence (C2) Vorticity (C3) U-component (C4) V-component

(d) Velocity magnitude (d1) Divergence (D2) Vorticity (D3) U-component (D4) V-component
Figure 27: PIV analysis images after 5 sec for 30 fps

40
After 10 sec from reference time:

(a) Velocity magnitude (A1) Divergence (A2) Vorticity (A3) U-component (A4) V-component

(b) Velocity magnitude (B1) Divergence (B2) Vorticity (B3) U-component (B4) V-component

41
(c) Velocity magnitude (C1) Divergence (C2) Vorticity (C3) U-component (C4) V-component

(d) Velocity magnitude (D1) Divergence (D2) Vorticity (D3) U-component (D4) V-component
Figure 28: PIV analysis images after 10 sec for 30 fps

42
 4.3.2. 60 fps (Start ignition t = 0):

For 60 fps video, calibration was done using following time and length that can be measured from background:

1
𝑡=
60

𝑡 = 16.667 ∗ 10−3 𝑠𝑒𝑐

𝑡 = 16.667 𝑚𝑠𝑒𝑐

Following are figures shown in sequence, first at t = 0 then after 5 and 10 sec of ignition to reveal progress of flame flow from
burner tip to top at 60 fps and flame structure.

43
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 29: PIV analysis images 1 at reference time for 60 fps

44
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 30: PIV analysis images 2 at reference time for 60 fps

45
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 31: PIV analysis images 3 at reference time for 60 fps

46
(a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) U-component (g) V-component (h) Difference image 1 (i) Difference image 2

Figure 32: PIV analysis images 4 at reference time for 60 fps

47
After 5 sec from reference time:

(a) Velocity magnitude (A1) Divergence (A2) Vorticity (A3) U-component (A4) V-component

(b) Velocity magnitude (B1) Divergence (B2) Vorticity (B3) U-component (B4) V-component

48
(c) Velocity magnitude (C1) Divergence (C2) Vorticity (C3) U-component (C4) V-component

(d) Velocity magnitude (D1) Divergence (D2) Vorticity (D3) U-component (D4) V-component

Figure 33: PIV analysis images after 5 sec for 60 fps

49
After 10 sec from reference time:

(a) Velocity magnitude (A1) Divergence (A2) Vorticity (A3) U-component (A4) V-component

(b) Velocity magnitude (B1) Divergence (B2) Vorticity (B3) U-component (B4) V-component

50
(c) Velocity magnitude (C1) Divergence (C2) Vorticity (C3) U-component (C4) V-component

(d) Velocity magnitude (D1) Divergence (D2) Vorticity (D3) U-component (D4) V-component

Figure 34: PIV analysis images after 10 sec for 60 fps

51
 4.3.3. 240 fps (Start ignition t = 0):

For 240 fps video, calibration was done using following time and length that can be measured from background:

1
𝑡=
240

𝑡 = 4.163 ∗ 10−3 𝑠𝑒𝑐

𝑡 = 4.163 𝑚𝑠𝑒𝑐

Following are figures shown in sequence, first at t = 0 then after 5 and 10 sec of ignition to display growth of flame flow from
start to top at 240 fps and flame structure.

52
 (a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) Difference image 1 (g) Difference image 2

Figure 35: PIV analysis images 1 at reference time for 240 fps

53
 (a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) Difference image 1 (g) Difference image 2

Figure 36: PIV analysis images 2 at reference time for 240 fps

54
 (a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) Difference image 1 (g) Difference image 2

Figure 37: PIV analysis images 3 at reference time for 240 fps

55
 (a) Raw images 1 (b) Raw images 2 (c) Velocity magnitude (d) Divergence (e) Vorticity

(f) Difference image 1 (g) Difference image 2

Figure 38: PIV analysis images 4 at reference time for 240 fps

56
After 5 sec from reference time:

(a) Velocity magnitude (A1) Divergence (A2) Vorticity

(b) Velocity magnitude (B1) Divergence (B2) Vorticity

57
(c) Velocity magnitude (C1) Divergence (C2) Vorticity

(d) Velocity magnitude (D1) Divergence (D2) Vorticity

Figure 39: PIV analysis images after 5 sec for 240 fps

58
After 10 sec from reference time:

(a) Velocity magnitude (A1) Divergence (A2) Vorticity

59
(b) Velocity magnitude (B1) Divergence (B2) Vorticity

(c) Velocity magnitude (C1) Divergence (C2) Vorticity

(d) Velocity magnitude (D1) Divergence (D2) Vorticity

Figure 40: PIV analysis images after 10 sec for 240 fps

60
 As stated above in the literature review, the structure of premixed flames was in
conical shape. It could be observed from all the figures and specially from vorticity and
absolute difference images (black and white) that “flame structure is conical” and
results shows that premixed flames only has single flame structure/zone. Flame zones
occur when more combustion takes place at the flame front as in the case of diffusion
flame. Length of the flame can be observed by the percentage of air in the burner; more
percentage of air is fed into burner, longer will be the structure of flame. Currently, inlets
air and fuel were not controlled. Air and fuel into inlets should be more controlled and it
will make it possible to study the effect of surrounding air i.e. by changing air percentage
and level of combustion completeness.

Figure 41: Effect of surrounding air and flame plume at stable position

Effect of surrounding air and flame plume can be observed in the Figure 4141. From the
figure above, it could be observed that flame propagate upwards in a axisymmetric way. The
flame structure was diffused by surrounding air, due to the effect of temperature difference,
as a jet which was coming in and starts to develop a vortex which was spreading out and
moving upward. In few seconds after ignition, flame attains a stable structure. But this
stability becomes irregular due to diffusion process. Flame displaced in the velocity range of
12 - 15 mm/sec. Air diffusion increases velocity and maximum velocity which has been
observed in this experiment is 20 mm/sec.

61
 As we increase number of frames per second (fps), we can get more information about
motion (i.e. more images per second), but down side of having high fps is lighting
requirement will increase significantly. There are also limitation to CMOS (Complementary
metal-oxide semiconductor) based camera sensor to handle capture and shutter, because
CMOS sensor works like a scanner. It scans down the image, which is called rolling shutter
(figure 42). If we need to really on 240 fps or higher frames per seconds, then we need to
compensate for resolution, varying background and shutter speed.

Figure 42: Effect of rolling shutter on moving propeller [37]

4.4. BOS experiment using mirror:

The PIV images are two-dimensional (2-D); therefore it is not possible to get 3-D images.
This study is not focused on 3-D images but concerns with multiple side-views using a mirror
because the cameras/lens are an expensive part of the experiment. Another experiment is done by
adding a mirror in the existing setup to get a different view (figure 43). The addition of the
mirror is to check the applicability of a mirror used in BOS experiment, or in other words “is it
possible to get results/images using a mirror”?

62
 Figure 43: Diagram depicting experimental setup using mirror

4.4.1. 30 frame per second:

Start ignition t = 0 (taken as reference for time):

Following are figures shown in sequence, first at t = 0 then after 5 and 10 sec of ignition to show
progress of flame flow from burner base to top at 30 fps and flame structure.

63
(a) Raw image 1 (b) Raw image 2 (c) Velocity magnitude (d) Divergence

(e) Vorticity (f) Difference image 1 (g) Difference image 2

Figure 44: PIV analysis images 1 using mirror at reference time for 30 fps

64
(a) Raw image 1 (b) Raw image 2 (c) Velocity magnitude (d) Divergence

(e) Vorticity (f) Difference image 1 (g) Difference image 2

Figure 45: PIV analysis images 2 using mirror at reference time for 30 fps

65
(a) Raw image 1 (b) Raw image 2 (c) Velocity magnitude (d) Divergence

(e) Vorticity (f) Difference image 1 (g) Difference image 2

Figure 46: PIV analysis images 3 using mirror at reference time for 30 fps

66
After 5 sec from reference time:

(a) Velocity magnitude (a1) Divergence (a2) Vorticity

(b) Velocity magnitude (b1) Divergence (b2) Vorticity

67
(c) Velocity magnitude (c1) Divergence (c2) Vorticity

(d) Velocity magnitude (d1) Divergence (d2) Vorticity

Figure 47: PIV analysis images using mirror after 5 sec for 30 fps

68
After 10 sec from reference time:

(a) Velocity magnitude (a1) Divergence (a2) Vorticity

(b) Velocity magnitude (b1) Divergence (b2) Vorticity

69
 (c) Velocity magnitude (c1) Divergence (c2) Vorticity

(d) Velocity magnitude (d1) Divergence (d2) Vorticity

Figure 48: PIV analysis images using mirror after 10 sec for 30 fps

4.4.2. 60 frames per second:

Start ignition t = 0 (taken as reference for time):

Following are figures shown in sequence, first at t = 0 then after 5 and 10 sec of ignition to
demonstration movement of flame flow from bottom to top at 60 fps and flame structure.

70
(a) Raw image 1 (b) Raw image 2 (c) Velocity magnitude (d) Divergence

(e) Vorticity (f) Difference image 1 (g) Difference image 2

Figure 49: PIV analysis images 1 using mirror at reference time for 60 fps

71
(a) Raw image 1 (b) Raw image 2 (c) Velocity magnitude (d) Divergence

(e) Vorticity (f) Difference image 1 (g) Difference image 2

Figure 50: PIV analysis images 2 using mirror at reference time for 60 fps

72
(a) Raw image 1 (b) Raw image 2 (c) Velocity magnitude (d) Divergence

(e) Vorticity (f) Difference image 1 (g) Difference image 2

Figure 51: PIV analysis images 3 using mirror at reference time for 60 fps

73
(a) Raw image 1 (b) Raw image 2 (c) Velocity magnitude (d) Divergence

(e) Vorticity (f) Difference image 1 (g) Difference image 2

Figure 52: PIV analysis images 4 using mirror at reference time for 60 fps

74
After 5 sec from reference time:

(a) Velocity magnitude (a1) Divergence (a2) Vorticity

(b) Velocity magnitude (b1) Divergence (b2) Vorticity

75
(c) Velocity magnitude (c1) Divergence (c2) Vorticity

(d) Velocity magnitude (d1) Divergence (d2) Vorticity

Figure 53: PIV analysis images using mirror after 5 sec for 60 fps

76
After 10 sec from reference time:

(a) Velocity magnitude (a1) Divergence (a2) Vorticity

(b) Velocity magnitude (b1) Divergence (b2) Vorticity

77
 (c) Velocity magnitude (c1) Divergence (c2) Vorticity

(d) Velocity magnitude (d1) Divergence (d2) Vorticity

Figure 54: PIV analysis images using mirror after 10 sec for 60 fps

Another experiment was done to get different view of flame propagation using mirror. Some of
the observation are listed below:

 Very light flame propagation through mirror could be observed but we had lost most of
intensity/energy between mirror image and normal image. One of the reason behind it is
the internal reflection of the mirror. Light goes into glass, some of the light get reflected
back to viewer but some part shifted slightly and get into transparent material of mirror.
When light hit coaster layer reflected back, due to internal reflection we get two images.

78
 As focus was on two backgrounds (one normal and other 90 degree flipped due to
mirror), so it is difficult to attain sharp focus on both backgrounds and also on subject.
 A better first surface mirror is required to prevent internal reflection or loss of light
intensity and to get sharp focus on schlieren object with backgrounds.

Figure 55: Internal reflection in mirror [38]

79
Chapter 5: Conclusion & Future Work

5.1. Conclusion:

In the present study, an experimental setup for studying premixed flames has been developed and
used for tests. The focus has been to study structure and propagation of flame. The images of
flame was acquired by using smartphone camera and the velocities of flame front computed
using PIVlab (MATLAB application) to find out how fast flame front propagates from bottom to
top. Moreover, a MATLAB program was developed which can subtract two images to
demonstrate flame structure.

The available camera frame rate is the limiting factor for smartphone background oriented
schlieren velocimetry. It should be sufficiently high that the imaged flow experiences a few
pixels displacement between frames. If fps (frame per second) rate increases, then more light is
need on the background as camera have less time to take an image (more frames in a second).
The camera uses CMOS sensor which works like a scanner, it scans down the chip to create
image. Down side of using CMOS sensor, if an object is moving in the same direction of scanner
(normally top to bottom) then sensor can stretched out the shape of the object. Inversely if the
object is moving in opposite direction, then it is going to compress that object.

Real flows are compressible, 3-dimensional, and unsteady in nature. The 3D info of flow can be
quantified by capturing data from multiple directions on the flow field. Effort was done to get
information from two different sides, one normal camera plane and the other 90 degree flipped
using mirror that was focus on another BOS background. Main purpose of using mirror is to
capture flame images from a side view, but some issues like loss of light intensity, internal
reflection (first surface mirror) needs to be addressed.

5.2. Future work:

At the end of this report, it is viable to point-out some potential future research areas. Following
are some workable future research areas:

80
 This method can be used to study effect of air percentage (if air inlet is controlled),
burner diameter and different fuel types on structure of flame.
 This technique can be used for gas leakage detection, gases which are colourless and
odourless.
 It is highly recommended to simulate premixed flame using CFD (Computational Fluid
Dynamic) modeling to compare the experimental results.
 Use of natural background to measure density variation at large scale.

81
References

[1] Roy, D. (1997). The cost of fires: a review of the information available, Home Office Publication
Unit, London.

[2] Drysdale, D. (1999). An introduction to fire dynamics, 2nd edition, pp. 1-130. John Wiley & Sons.

[3] Warnatz, J., Mass, U., Dibble, R.W. (2006). Combustion: physical and chemical fundamentals,
modelling and simulation, experiments, pollutant formation, 4th edition, Springer-Verlag Berlin
Heidelberg.

[4] Sevault, A. (2012). Investigation of the Structure of Oxy-Fuel Flames Using Raman Laser
Diagnostics. Doctoral Theses, NTNU, https://brage.bibsys.no/xmlui/handle/11250/235022 [Accessed 2
Jan. 2019].

[5] YouTube. (2019). Premixed Flame vs Diffusion Flame - Fire Protection Engineering (FPE) Teaching
Tool. [Online] Available at: https://www.youtube.com/watch?v=_gqO3ncLfRg [Accessed 30 May 2019].

[6] Hottel, H.C., and Hawthorne, W.R. (1949). 3rd Symposium (International) on Combustion, pp. 255-
266. Williams and Wilkins, Baltimore.

[7] Wang, Q. (2012). Advanced Optical and 3D Reconstruction Diagnostics for Combustion and Fluids
Research, pp. 30-55, PhD thesis, University of Sheffield.

[8] Richard, H., Raffel, M., Rein, M., Kompenhans, J. and Meier, G. (2002). Demonstration of the
applicability of a Background Oriented Schlieren (BOS) method, Laser Techniques for Fluid Mechanics,
pp.145-156.

[9] Tokgoz, S., R. Geisler, L. J. A. van Bokhoven and B. Wieneke (2012). Temperature and velocity
measurements in a fluid layer using background-oriented schlieren and PIV methods, Measurement
Science and Technology 23(11): 115302.

[10] Richard, H., Raffel, M. (2001). Principle and applications of the background oriented schlieren
(BOS) method, Measurement Science and Technology 12 1576–85.

[11] Meier, G.E.A. (2002). Computerized background-oriented schlieren, Exp. Fluids 33 181–7.

82
[12] Dalziel, S.B., Hughes, G.O., Sutherland, B.R. (2000). Whole-field density measurements by
‘synthetic schlieren’, Exp. Fluids 28 322–35.

[13] Venkatakrishnan, L., Meier, G.E.A. (2004). Density measurements using the background oriented
schlieren technique, Exp. Fluids 37 237–47.

[14] Merzkirch W., Egami Y. (2007). Density-Based Techniques. In: Tropea C., Yarin A.L., Foss J.F.
(eds) Springer Handbook of Experimental Fluid Mechanics. Springer Handbooks. Springer, Berlin,
Heidelberg.

[15] Elsinga, G.E., van Oudheusden, B.W., Scarano, F., Watt, D.W. (2004). Assessment and
application of quantitative schlieren methods: calibrated color schlieren and background oriented
schlieren, Exp. Fluids 36 309–25.

[16] Goldhahn, E., Seume, J. (2007). The background oriented schlieren technique: sensitivity, accuracy,
resolution and application to a three-dimensional density field, Exp. Fluid 43 241–9.

[17] Atcheson, B., Heidrich, W., Ihrke, I. (2008). An evaluation of optical flow algorithms for
background oriented schlieren imaging, Exp. Fluids 46 467–76.

[18] Hargather, M.J., Settles, G.S. (2010). Natural-background oriented schlieren imaging, Exp. Fluids 48
59–68.

[19] Venkatakrishnan, L., Wiley, A., Kumar, R., Alvi, F. (2011). Density field measurements of a
supersonic impinging jet with microjet control, AIAA Journal 49 432–8.

[20] Settles, G.S. (2018). Smartphone schlieren and shadowgraph imaging, Optics and Lasers in
Engineering 104: 9-21.

[21] Hill, M. and Haering, E. (2016). Ground-to-air flow visualization using Solar Calcium-K line
Background-Oriented Schlieren, Experiments in Fluids, 58(1).

[22] Settles, G.S., Hargather M.J. (2017). A review of recent developments in schlieren and shadowgraph
techniques, Measurement Science and Technology 28(4): 042001.

[23] Raffel, M. (2015). Background-oriented schlieren (BOS) techniques, Exp. Fluids 56 60.

83
[24] Friel, J., Grande, J. and Hetzner, D. (2000). Practical guide to image analysis, Materials Park (Ohio):
ASM International.

[25] Russ, J. (2011). The image processing handbook, Bosa Roca: CRC Press.

[26] Woynar, L. (1998). Image Analysis: Applications in Materials Engineering, Boca Raton, Fla.: CRC
Press.

[27] Ce, L., W. T. Freeman, R. Szeliski and K. Sing Bing (2006). Noise Estimation from a Single Image,
2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06).

[28] Thielicke, W., Stamhuis, E.J. (2014). PIVlab – Towards User – friendly, Affordable and Accurate
Digital Particle Image Velocimetry in MATLAB, Journal of open research software, 2(E30).
DOI: http://doi.org/10.5334/jors.bl

[29] Raffel, M., Willert, C., Wereley, S. and Kompenhans, J. (2007). Particle Image Velocimetry,
Dordrecht: Springer.

[30] Shavit, U., Lowe, R. and Steinbuck, J. (2006). Intensity Capping: a simple method to improve cross-
correlation PIV results, Experiments in Fluids, 42(2), pp.225-240.

[31] Pizer, S. M., E. P. Amburn, J. D. Austin, R. Cromartie, A. Geselowitz, T. Greer, B. ter Haar
Romeny, J. B. Zimmerman and K. Zuiderveld (1987). Adaptive histogram equalization and its variations,
Computer Vision, Graphics, and Image Processing 39(3): 355-368.

[32] Otsuka, T., Wolanski, P. (2001). Particle image velocimetry (PIV) analysis of flame structure,
Journal of Loss Prevention in the Process Industries, 14(1), pp. 503-507.

[33] Yin, R., L. Xu and X. Li (2015). A simplified PIV-based method for flame velocity distribution
measurement, 2015 IEEE International Conference on Imaging Systems and Techniques (IST).

[34] F. Moisy, PIVMat, A PIV post-processing and data analysis toolbox for MATLAB, http://www.fast.u-
psud.fr/pivmat/ [Accessed 15 Jan. 2019].

[35] https://www.mathworks.com/matlabcentral/fileexchange/27659-pivlab-particle-image-velocimetry-
piv-tool [Accessed 10 Jan. 2019].

84
[36] Schneider, C.A., Rasband, W.S., Eliceiri, K.W. NIH image to ImageJ: 25 years of image analysis.
Nat Meth 2012; 9:671-5 See also https://imagej.net/ImageJ [Accessed 2 Jan. 2019].

[37] YouTube. (2019). Rolling Shutter Explained (Why Do Cameras Do This?) - Smarter Every Day 172.
[Online] Available at: https://www.youtube.com/watch?v=dNVtMmLlnoE [Accessed 30 May 2019].

[38] En.wikibooks.org. (2019). A-level Physics (Advancing Physics)/what is a wave? /Worked Solutions
- Wikibooks, open books for an open world. [Online] Available at: https://en.wikibooks.org/wiki/A-
level_Physics_(Advancing_Physics)/What_is_a_wave%3F/Worked_Solutions [Accessed 24 April 2019].

[39] Yang, W. (2001). Handbook of flow visualization. New York: Taylor and Francis.

[40] FutureLearn, (2019). Image analysis methods for biologists. [online] Available at:
https://www.futurelearn.com/courses/image-analysis [Accessed 3 Jan. 2019].

[41] En.wikipedia.org. (2019). Divergence. [online] Available at:

https://en.wikipedia.org/wiki/Divergence [Accessed 13 Mar. 2019].

[42] Betterexplained.com. (2019). Vector Calculus: Understanding Divergence – Better Explained.

[online] Available at: https://betterexplained.com/articles/divergence/ [Accessed 13 Mar. 2019].

85
Appendix A

MATLAB script for BOS dot pattern [34]:

“function makebospattern(n,diam,figtype,filename)”

“%MAKEBOSPATTERN Random dot pattern for SS / BOS / FS-SS applications”

“% MAKEBOSPATTERN(N,D) makes a figure filled with N randomly distributed”

“% dots of diameter D for SS (Synthetic Schlieren), BOS (Background-”

“% oriented Schlieren) or FS-SS (Free-Surface SS) applications.”

“%”

“% The figure format is portrait (ie, vertical) A4, 210x297 mm, and the”

“% particle diameter D is in mm. Typical values are N=50000, D=1.”

“%”

“% MAKEBOSPATTERN(N,D,'b') makes black points on a white ground (by”

“% default). MAKEBOSPATTERN(N,D,'w') makes white points on a black ground.”

“%”

“% MAKEBOSPATTERN(N,D,'..',FILENAME) saves the result in a 300-DPI TIFF”

“% figure. Use the standard Windows viewer to print the figure.”

“%”

“% Notes:”

“% - If you zoom the figure, the particles are not resized: particles”

86
“% are drawn in absolute units (points), while the paper size is in”

“% physical units (centimeters).”

“% - a 'particle' is a set of 4 concentric circles with increasing (or”

“% decreasing) gray levels. The 'particle diameter' D is the diameter of”

“% the outer circle; the inner circle has diameter 0.3*D. Depending on”

“% the printer quality, the outer circle may not print correctly, so the”

“% actual size may appear slightly larger or smaller than the requested”

“% size.”

“% - for FS-SS applications, see the function SURFHEIGHT for the”

“% surface height reconstruction.”

“%”

“% Example:”

“% makebospattern(50000,1,'w','myfig');”

“%”

“% See also PRINT, SURFHEIGHT”

“% F. Moisy, moisy_at_fast.u-psud.fr”

“% Revision: 1.21, Date: 2009/04/14.”

87
“% This function is part of the PIVMat Toolbox”

“% History:”

“% 2006/06/26: v1.00, first version ('makepidpattern')”

“% 2006/07/03: v1.10, S in mm; black/white option; saves in a TIFF file.”

“% 2006/07/05: v1.11, no need to set the PaperPosition and Background mode.”

“% 2007/05/29: v1.12, minor additional statistics”

“% 2008/09/05: v1.20, now entitled 'makebospattern'; included in PIVMat.”

“% 2009/04/14: v1.21, help text changed (BOS->SS)”

“% default values:”

“if ~exist('n','var'), n=50000; end”

“if ~exist('diam','var'), diam=0.25; end”

“if ~exist('figtype','var'), figtype='b'; end”

“h=gcf;”

88
“% Window size:”

“set(h,'Position',[360 80 560/sqrt(2) 560]);”

“set(gca,'Position',[0 0 1 1]); % location of the figure in the window”

“set(gca,'PlotBoxAspectRatio',[1/sqrt(2) 1 1]);”

“% Printing settings:”

“set(h,'PaperUnits','centimeters');”

“set(h,'PaperOrientation','portrait');”

“set(h,'PaperType','A4');”

“%set(h,'PaperSize',[21 29.7]);”

“set(h,'PaperPosition',[0 0 21 29.7]);”

“set(h,'PaperPositionMode','manual');”

“set(h,'InvertHardcopy','off'); % keep the user background mode”

“switch lower(figtype),”

“ case 'b' % black points on a white ground”

“ col{1} = 0.8*ones(1,3);”

“ col{2} = 0.6*ones(1,3);”

“ col{3} = 0.4*ones(1,3);”

89
“ col{4} = 0 * ones(1,3);”

“ set(h,'Color',[1 1 1]);”

“ case 'w' % white points on a black ground”

“ col{1} = 0.4*ones(1,3);”

“ col{2} = 0.6*ones(1,3);”

“ col{3} = 0.8*ones(1,3);”

“ col{4} = 1 * ones(1,3);”

“ set(h,'Color',[0 0 0]);”

“end”

“x=rand(1,n);”

“y=rand(1,n);”

“s = diam * 72/25.4; % diameter, in points units (1 point = 1/72 inch = 25.4/72 mm)”

“plot(x,y,'o','MarkerFaceColor',col{1},'MarkerEdgeColor',col{1},'Markersize',s);”

“hold on”

“plot(x,y,'o','MarkerFaceColor',col{2},'MarkerEdgeColor',col{2},'Markersize',0.7*s);”

“plot(x,y,'o','MarkerFaceColor',col{3},'MarkerEdgeColor',col{3},'Markersize',0.5*s);”

“plot(x,y,'o','MarkerFaceColor',col{4},'MarkerEdgeColor',col{4},'Markersize',0.3*s);”

90
“hold off”

“axis off;”

“%display some info:”

“disp(' ');”

“disp('PID pattern parameters:');”

“switch lower(figtype)”

“ case 'b'”

“ disp([' ' num2str(n) ' black particles on a 210x297 mm (A4) paper']);”

“ case 'w'”

“ disp([' ' num2str(n) ' white particles on a 210x297 mm (A4) paper']);”

“end”

“disp([' Particle diameter = ' num2str(diam) ' mm.']);”

“disp([' ' num2str(n/(210*297)) ' particles / mm^2']);”

“disp([' filled surface ratio = ' num2str((n*pi*diam^2/4)/(297*210))]);”

“disp(' ');”

“disp('Assuming a 1280x1024 camera, with 210 mm = 1024 pixels:');”

“pdiam = diam/210*1024;”

91
“disp([' particle diameter = ' num2str(pdiam) ' pixels']);”

“if pdiam<1.3”

“ disp(' * Warning: particles too small *');”

“elseif pdiam>4”

“ disp(' * Warning: particles too big *');”

“end”

“ncam = n * (5/4) / sqrt(2); % number of particles in a 5/4 camera field (eg, 1280x1024).”

“ppp = ncam / (1280*1024);”

“disp([' ' num2str(ppp) ' particles / pixel^2']);”

“if ppp>0.2”

“ disp(' * Warning: high density *');”

“elseif ppp<0.02”

“ disp(' * Warning: low density *');”

“end”

“disp(' ');”

“disp('Particles per interrogation window:');”

“winsize = [6 8 12 16 32 64];”

“for nws=1:length(winsize)”

92
“ ppw (nws) = winsize(nws)^2 * ncam / (1280*1024);”

“ disp([' ' num2str(ppw (nws)) ' particles / ' num2str(winsize(nws)) '^2-window']);”

“end”

“optws = sqrt(5*(1280*1024)/ncam);”

“disp([' Optimal window size: ' num2str(optws,3) '^2']);”

“dif=abs(ppw-5);”

“ind=find(dif==min(dif));”

“disp([' Closest window size: ' num2str(winsize(ind)) '^2']);”

“disp(' ');”

“% save the figure:”

“if exist('filename','var')”

“ print('-dtiff','-r500',filename);”

“ disp(['''' filename ''' saved']);”

“end”

93
Appendix B

MATLAB script for seeding particles with dot pattern [34]:

“%function makebospattern(n,diam,figtype,filename)”

“%MAKEBOSPATTERN Random dot pattern for SS / BOS / FS-SS applications”

“% MAKEBOSPATTERN(N,D) makes a figure filled with N randomly distributed”

“% dots of diameter D for SS (Synthetic Schlieren), BOS (Background-”

“% oriented Schlieren) or FS-SS (Free-Surface SS) applications.”

“%”

“% The figure format is portrait (ie, vertical) A4, 210x297 mm, and the”

“% particle diameter D is in mm. Typical values are N=50000, D=1.”

“%”

“% MAKEBOSPATTERN(N,D,'b') makes black points on a white ground (by”

“% default). MAKEBOSPATTERN(N,D,'w') makes white points on a black ground.”

“%”

“% MAKEBOSPATTERN(N,D,'..',FILENAME) saves the result in a 300-DPI TIFF”

“% figure. Use the standard Windows viewer to print the figure.”

“%”

“% Notes:”

94
“% - If you zoom the figure, the particles are not resized: particles”

“% are drawn in absolute units (points), while the paper size is in”

“% physical units (centimeters).”

“% - a 'particle' is a set of 4 concentric circles with increasing (or”

“% decreasing) gray levels. The 'particle diameter' D is the diameter of”

“% the outer circle; the inner circle has diameter 0.3*D. Depending on”

“% the printer quality, the outer circle may not print correctly, so the”

“% actual size may appear slightly larger or smaller than the requested”

“% size.”

“% - for FS-SS applications, see the function SURFHEIGHT for the”

“% surface height reconstruction.”

“%”

“% Example:”

“% makebospattern(50000,1,'w','myfig');”

“%”

“% See also PRINT, SURFHEIGHT”

“% F. Moisy, moisy_at_fast.u-psud.fr”

95
“% Revision: 1.21, Date: 2009/04/14.”

“% This function is part of the PIVMat Toolbox”

“% History:”

“% 2006/06/26: v1.00, first version ('makepidpattern')”

“% 2006/07/03: v1.10, S in mm; black/white option; saves in a TIFF file.”

“% 2006/07/05: v1.11, no need to set the PaperPosition and Background mode.”

“% 2007/05/29: v1.12, minor additional statistics”

“% 2008/09/05: v1.20, now entitled 'makebospattern'; included in PIVMat.”

“% 2009/04/14: v1.21, help text changed (BOS->SS)”

“clc”

“clear all”

“% default values:”

“if ~exist('n','var'), n=5000; end”

“%if ~exist('diam','var'), diam=0.25; end”

“if ~exist('diam','var'), diam=1; end”

96
“if ~exist('figtype','var'), figtype='b'; end”

“n”

“h=gcf;”

“% Window size:”

“set(h,'Position',[360 80 560/sqrt(2) 560]);”

“set(gca,'Position',[0 0 1 1]); % location of the figure in the window”

“set(gca,'PlotBoxAspectRatio',[1/sqrt(2) 1 1]);”

“% Printing settings:”

“set(h,'PaperUnits','centimeters');”

“set(h,'PaperOrientation','portrait');”

“set(h,'PaperType','A4');”

“%set(h,'PaperSize',[21 29.7]);”

“set(h,'PaperPosition',[0 0 21 29.7]);”

“set(h,'PaperPositionMode','manual');”

“set(h,'InvertHardcopy','off'); % keep the user background mode”

“switch lower(figtype),”

97
“ case 'b' % black points on a white ground”

“ col{1} = 0.8*ones(1,3);”

“ col{2} = 0.6*ones(1,3);”

“ col{3} = 0.4*ones(1,3);”

“ col{4} = 0 * ones(1,3);”

“ set(h,'Color',[1 1 1]);”

“ case 'w' % white points on a black ground”

“ col{1} = 0.4*ones(1,3);”

“ col{2} = 0.6*ones(1,3);”

“ col{3} = 0.8*ones(1,3);”

“ col{4} = 1 * ones(1,3);”

“ set(h,'Color',[0 0 0]);”

“end”

“x=rand(1,n);”

“y=rand(1,n);”

“%Seeding particles:”

“xsA = rand(1,1000);”

98
“ysA = rand(1,1000)*0.9;”

“%Shift seeding particles A along the y-axis:”

“xsB = xsA;”

“ysB = ysA + 0.01;”

“s = diam * 72/25.4; % diameter, in points units (1 point = 1/72 inch = 25.4/72 mm)”

“figure(1)”

“plot(x,y,'o','MarkerFaceColor','k','MarkerEdgeColor','k','Markersize',s);”

“% plot(x,y,'o','MarkerFaceColor',col{1},'MarkerEdgeColor',col{1},'Markersize',s);”

“% hold on”

“% plot(x,y,'o','MarkerFaceColor',col{2},'MarkerEdgeColor',col{2},'Markersize',0.7*s);”

“% plot(x,y,'o','MarkerFaceColor',col{3},'MarkerEdgeColor',col{3},'Markersize',0.5*s);”

“% plot(x,y,'o','MarkerFaceColor',col{4},'MarkerEdgeColor',col{4},'Markersize',0.3*s);”

“% hold off”

“axis off;”

“% ******************”

“% Window size:”

99
“figure(2)”

“h2=gcf;”

“set(h2,'Position',[360 80 560/sqrt(2) 560]);”

“set(gca,'Position',[0 0 1 1]); % location of the figure in the window”

“set(gca,'PlotBoxAspectRatio',[1/sqrt(2) 1 1]);”

“hold on”

“plot(x,y,'o','MarkerFaceColor','k','MarkerEdgeColor','k','Markersize',s);”

“plot(xsA,ysA,'o','MarkerFaceColor','r','MarkerEdgeColor','r','Markersize',s);”

“hold off”

“axis off;”

“% ******************”

“figure(3)”

“h3=gcf;”

“set(h3,'Position',[360 80 560/sqrt(2) 560]);”

“set(gca,'Position',[0 0 1 1]); % location of the figure in the window”

“set(gca,'PlotBoxAspectRatio',[1/sqrt(2) 1 1]);”

“hold on”

“plot(x,y,'o','MarkerFaceColor','k','MarkerEdgeColor','k','Markersize',s);”

“plot(xsB,ysB,'o','MarkerFaceColor','r','MarkerEdgeColor','r','Markersize',s);”

100
“hold off”

“axis off;”

“%display some info:”

“disp(' ');”

“disp('PID pattern parameters:');”

“switch lower(figtype)”

“ case 'b'”

“ disp([' ' num2str(n) ' black particles on a 210x297 mm (A4) paper']);”

“ case 'w'”

“ disp([' ' num2str(n) ' white particles on a 210x297 mm (A4) paper']);”

“end”

“disp([' Particle diameter = ' num2str(diam) ' mm.']);”

“disp([' ' num2str(n/(210*297)) ' particles / mm^2']);”

“disp([' filled surface ratio = ' num2str((n*pi*diam^2/4)/(297*210))]);”

“disp(' ');”

“disp('Assuming a 1280x1024 camera, with 210 mm = 1024 pixels:');”

“pdiam = diam/210*1024;”

“disp([' particle diameter = ' num2str(pdiam) ' pixels']);”

101
“if pdiam<1.3”

“ disp(' * Warning: particles too small *');”

“elseif pdiam>4”

“ disp(' * Warning: particles too big *');”

“end”

“ncam = n * (5/4) / sqrt(2); % number of particles in a 5/4 camera field (eg, 1280x1024).”

“ppp = ncam / (1280*1024);”

“disp([' ' num2str(ppp) ' particles / pixel^2']);”

“if ppp>0.2”

“ disp(' * Warning: high density *');”

“elseif ppp<0.02”

“ disp(' * Warning: low density *');”

“end”

“disp(' ');”

“disp('Particles per interrogation window:');”

“winsize = [6 8 12 16 32 64];”

“for nws=1:length(winsize)”

“ ppw (nws) = winsize(nws)^2 * ncam / (1280*1024);”

102
“ disp([' ' num2str(ppw (nws)) ' particles / ' num2str(winsize(nws)) '^2-window']);”

“end”

“optws = sqrt(5*(1280*1024)/ncam);”

“disp([' Optimal window size: ' num2str(optws,3) '^2']);”

“dif=abs(ppw-5);”

“ind=find(dif==min(dif));”

“disp([' Closest window size: ' num2str(winsize(ind)) '^2']);”

“disp(' ');”

“% save the figure:”

“if exist('filename','var')”

“ print('-dtiff','-r500',filename);”

“ disp(['''' filename ''' saved']);”

“end”

103
Appendix C

MATLAB script for absolute difference images:

%analyse images by taking difference of any image by the reference

clear; close all; clc

I = imread('r.bmp');

%images upto 52 is at start of flame and then after 10 sec upto 106

for i=1:41

filn = [int2str(i) '.bmp']

%J = imread('40.bmp');

J = imread(filn);

K = imabsdiff(I,J)*8;

figure(1), imshow(K)

pause(0.2)

end

K2=K(129:700,75:400)

figure(2)

104
imshow(K2)

105
Appendix D

MATLAB script for video split:

“%%Extracting & Saving of frames from a Video file through Matlab Code%%”

“clc;”

“close all;”

“clear all;”

“% assigning the name of sample avi file to a variable”

“filename = 'slow video 1.mp4';”

“%reading a video file”

“mov = VideoReader(filename);”

“% Defining Output folder as 'snaps'”

“opFolder = fullfile(cd, 'snaps');”

“%if not existing”

“if ~exist(opFolder, 'dir')”

“%make directory & execute as indicated in opfolder variable”

“mkdir(opFolder);”

106
“end”

“%getting no of frames”

“numFrames = mov.NumberOfFrames;”

“%setting current status of number of frames written to zero”

“numFramesWritten = 0;”

“%for loop to traverse & process from frame '1' to 'last' frames”

“for t = 1 : numFrames”

“currFrame = read(mov, t); %reading individual frames”

“opBaseFileName = sprintf('%3.3d.bmp', t);”

“opFullFileName = fullfile(opFolder, opBaseFileName);”

“imwrite(currFrame, opFullFileName, 'bmp'); %saving as 'png' file”

“%indicating the current progress of the file/frame written”

“progIndication = sprintf('Wrote frame %4d of %d.', t, numFrames);”

“disp(progIndication);”

“numFramesWritten = numFramesWritten + 1;”

“end %end of 'for' loop”

107
“progIndication = sprintf('Wrote %d frames to folder "%s"',numFramesWritten, opFolder);”

“disp(progIndication);”

“%End of the code”

108