OBJECT COUNTING AND DENSITY CALCULATION

USING MATLAB

Submitted by

PREM KUMAR. V, BARATH. V, PRASHANTH. K

For more information contact: premkumarbullets@gmail.com,

prem91914@gmail.com

ABSTRACT
Counting the number of objects is an integral part of image processing.
Knowing the number of objects present in the image can be useful for further
analysis in a wide range of applications. In this project we propose a simple
method for automatically determining the number of objects in an image . Once
the number of objects are determined the objects per unit area or the density can
also be estimated. Existing methods involve counting based on area of objects,
color of objects, applying edge detection techniques etc. There are also
applications that involve a large amount of hardware components for counting
that further adds to the cost and maintenance which is a tedious work. The
proposed system involves converting the input image into a format such that
the number of objects can be calculated based on the connected components
present in the enhanced image. This project work also aims at determining the
correct value of density by clearing the objects touching the borders of the
image. In this project three applications are taken into account and using Matlab
with image processing toolbox the count and density values are calculated for
each.

CHAPTER 1
1.INTRODUCTION
Images contain a wealth of objects and patterns which may convey
information about underlying mechanism and methods. Counting involves
estimating the number of objects in an image or a still video frame. Counting
arises in many real-time applications such as counting cells in microscopic
images, number of people, number of vehicles on road etc.
This goal is to accurately estimate the count, but however we evade the
hard task of localizing or drawing boundary around the objects. Object detection
is a very hard problem to be solved, especially in case of overlapping objects.
Once the count is known , the density of objects or the number of objects
per unit area can be estimated.
As mentioned earlier, the existing works under counting involve a large
amount of hardware which also adds to the cost. Now with a simple image
processing software such as matlab many parameters can be estimated from
images. In this paper we have dealt with 3 such applications namely Cell counting
People counting based on faces
Vehicle counting
1.1 Cell counting:

COUNTING cells is often a necessary but tedious step for in vitro cell
culture.

Consistent cell concentrations ensure experimental reproducibility

and accuracy. Cell counts are important for monitoring cell health and
proliferation rate, assessing immortalization or transformation, seeding cells for
subsequent experiments, transfect ion or infection, and
Preparing for cell-based assays.
3

We propose a system for counting cells in image by using matlab and image
processing toolbox. It can be used for accurate calculation of object and density
of image
1.2 Counting people based on the faces detected:

One of the important applications of counting includes counting people

say in a hall or in a shopping mall etc,. People can be counted based on the
number of faces detected. Face detection is the process in which the human
faces are detected from a color image. Once the number of faces are detected
the density can be calculated. There are many ways by which faces can be
detected one such method is skin based detection. Rectangles can be used to
map the faces in the image. Nowadays Face detection is used in many day to
day application.

1.3 Vehicle counting:

Transportation research involves counting number of vehicles on road as

well as finding the density of traffic in a particular area. There are many
methods of detecting vehicles on road such as motion detection, installing lasers
on both sides of the road, etc., which is tedious and involves large number of
hardware. This method uses image processing techniques to count the number
of vehicles on road and estimate the density. The number of vehicles found can
be used for surveying or controlling the traffic signal.

1.4 EXISTING SYSTEMS:

1.4.1 CELL COUNT:
Methods include counting based on the area of objects, color of
objects, applying edge detection techniques etc.
1.4.2 Counting people based on the faces:

Existing methods based on skin detection provide inaccurate

count.

Existing method involves complex algorithms and difficult to

compute the faces.
Existing method of people counting involves a lot of hardware
components.
1.4.3 VEHICLE COUNT:
In existing system a timer based traffic operation is used which
does not estimate traffic parameters. This results in green light
being shown to an empty road.
Various sensors have been employed to estimate traffic parameters
for updating traffic information, magnetic loop detectors have been
the most used technologies in the existing system.
1.4.4 DEMERITS OF THE EXISTING SYSTEM:
Existing methods involve a large amount of hardware components.
Maintenance and cost are inconvenient.
Calculating the number of objects per unit area are not accurate.
The installation and maintenance of vehicle counting systems are very
high.

1.5 PROPOSED SYSTEM:

1.5.1 CELL COUNTING:
Counting cells in Drosophila is a complex task, due to variability in
image quality resulting from different cell markers. Cells are segmented
according to their characteristics. But cell shape changes with cell state
(i.e. arrest, mitosis, or apoptosis). For instance, during mitosis the shape is
irregular and it can be difficult to determine when a dividing cell can be
considered as two daughter cells. Nuclei and glia cells have a more regular
shape, between elliptical and circular. Apoptotic cells have initially a very
irregular shape, later on very round, and can appear subdivided into
different parts depending on the timing within apoptosis. Depending on the
kind of cells or cell state to be visualised, a different cell marker (i.e.
antibody) is employed. As a result, different image-processing methods
must be developed to quantify cells of different qualities.
Formula used:
OBJECTS PER UNIT AREA =

NO OF OBJECTS
TOTAL SIZE OF IMAGE

1.5.2 PEOPLE COUNTING:

In this proposed method we use skin based face detection for
counting people.
In this method multiple faces can be detected at a time or even a
group of people can be detected.
This method not only detects the faces but also count them.
From count the density can also be derived.
Both image and video can be given as input to the compiler.

1.5.3 VEHICLE COUNTING:

This proposed system of counting vehicles involves comparing the
images of road with and without vehicles.
On comparing two images we can find the difference between
images.
Based on the difference obtained, the number of vehicles as well as
count can be estimated.

1.6 SOFTWARE REQUIREMENTS:

MATLAB R2010a with image processing toolbox .

CHAPTER 2
2. LITERATURE SURVEY:
2.1 CELL COUNT:
Victor Lempitsky and Andrew Zisserman-Learning To Count
Objects in Images deals with the counting problem is the estimation of
the number of objects in a still image or video frame. It arises in many
real-world applications including cell counting in microscopic images,
monitoring crowds in surveillance systems, and performing wildlife
census or counting the number of trees in an aerial image of a forest. We
take a supervised learning approach to this problem, and so require a set
of training images with annotation.

Christian Wolf and Jean-Michel Jolion of Technical Report LIRIS

discussed about Object count/Area Graphs for the Evaluation of Object
Detection and Segmentation Algorithms Evaluation of object detection
algorithms is a non-trivial task: a detection result is usually evaluated by
comparing the bounding box of the detected object with the bounding
box of the ground truth object.

Andre R.S. Marcal Faculdade de Ciencias, Universidade do

Porto DMA-Alternative Methods for Counting Overlapping Grains in
Digital Images deals with Standard granulometry methods are used to
count the number of disjoint grains in digital images

2.2 PEOPLE COUNT:

Bhaskar Gupta , Sushant Gupta , Arun Kumar Tiwari of ABES
Engineering College, Ghaziabad discussed about Face Detection using
Gabor Feature and Artificial Neural Network which uses a Gabor filter
feature to extract facial features from the local image.

Sanjay Kr. Singh1, D. S. Chauhan2, Mayank Vatsa3, Richa Singh3

of Engineering Institute of Engineering and Technology Jaunpur, India
discussed about a Robust Skin Color Based Face Detection Algorithm.
They have compared the algorithms based on the color spaces
RGB,YCbCr and HSI and have combined them to get a new skin color
based face detection algorithm which gives higher accuracy.

Gary Chern, Paul Gurney, and Jared Starman discussed about Face
detection using Color based masking. They developed an algorithm
capable of locating each face in a color image of the class. They were
given seven training images along with the corresponding ground truth
data to develop and train algorithms on .

In 2003 Waqar Mohsin, Noman Ahmed, Chung-Tse Mar of

Stanford university discussed about an method to detect faces in an color
image using rejection based classification.

2.3 VEHICLE COUNT:

From the proceedings of the 26th National IT conference
,Colombo,Sri Lanka M. Pancharatnam, D.U.J. Sonnadara of Department
of physics, University of Colombo, Sri Lanka discussed about Vehicle
Counting and Classification from a Traffic Scene using adoptive
background subtraction technique. After the background subtraction,
using threshold and median filters, isolated image blobs are identified as
individual vehicles.
Erhan INCE from Department of Electrical and Electronic
Engineering, Eastern Mediterranean University, Famagusta, North
Cyprus, via Mersin 10 TURKEY discussed about Measuring traffic flow
and classifying vehicle types: A surveillance video based approach
based on invariant moments and shadow aware foreground masks.

Pejman Niksaz Science &Research Branch, Azad University of

Yazd, Iran discussed about Automatic Traffic Estimation Using Image
Processing based on background elimination and lane masking .
From the paper of L.G.C Wimalaratna1 and D.U.J. Sonnadara2,
1

Virtusa (Pvt.) Ltd., Trans Asia Commercial Complex, Colombo 2,

Department of Physics, University of Colombo, Colombo 3 which deals

with the Estimation of the Speeds of Moving Vehicles from Video

Sequences the concept of pre-processing video images using gray scale
and median filter to extract moving vehicles from a traffic scene is
surveyed. This method finds the difference between three consecutive
video frame and filters through edge detection filter Blob counting.
10

CHAPTER 3
3. BLOCK DIAGRAM:

3.1 CELL COUNT:

IDENTIFYING
BORDER
TOUCHING OBJECT

CONVERT COLOR TO
BINARY

INPUT IMAGE

COUNTING REMAINING
OBJECTS & CALCULATING
DENSITY

REMOVING BORDER
TOUCHING OBJECTS

Fig 3.1 Block diagram of the cell count

EXPLANATION:
First the input image is given to matlab, using imread()
function the image is converted from color to binary. Then the objects
touching the borders are identified and removed. The remaining objects
are counted and the density is carried out.
3.2 PEOPLE COUNT:

CONVERSION FROM
RGB TO LAB

INPUT IMAGE

DISPLAY NO. OF
PEOPLE AND DENSITY

APPLY GRAY THRESHOLD

AND EQUALISE
BRIGHTNESS

SHOW BLOBS AREA >1500

Fig 3.2 Block diagram of the people count

11

CONVERT TO BINARY

EXPLANATION:
First the input image is fed into matlab. The image is
converted from rgb to lab. Then the gray threshold is applied to the
converted image and brightness equalization is done. Then the equalized
image is converted to binary. The blobs in the image are opened when the
blobs area is greater than 1500. The number of people and density are
displayed.

3.3 VEHICLE COUNTING:

INPUT IMAGE WITH

EMPTY ROADS

CONVERTING FROM
RGB TO GRAY

COMPARE TWO IMAGES

AND FIND DIFFERENCE

INPUT IMAGE WITH

CARS ON ROAD

DISPLAY COUNT
AND DENSITY

FIND BLOBS OF
AREA >4000

CONVERT TO
BINARY

Fig 3.3 Block diagram of the vehicle count

EXPLANATION:
In vehicle counting two input images are given, one without cars
and the other with cars in it. The input image is then converted from rgb to gray.
Now compare the two images and find the difference. Then the image is
converted to binary. The blobs in the image are opened when the blob area is
greater than 4000. Finally the count and the density are displayed.

12

CHAPTER 4

4. ALGORITHM:
4.1 CELL COUNT:
Step 1)Start the program
Step 2)Get the input image using imread ()
Step 3)Convert color to gray if it is a color image.
Step 4)Use clear borders command if any object is touching the
border.
Step 5)Choose an appropriate threshold value using graythresh ()
command
Step 6)Fill blobs that occur with holes
Step 7)Apply morphological operation by using a filter of ones
Step 8)Display all blobs that have an area>100
Step 9)Count them using conncomp () command
Step 10) Calculate the density
Step 11) Stop the program

13

4.2 PEOPLE COUNT :

Step 1: Start the program
Step 2: Get the input image
Step 3: Change the color space from rgb to lab
Step 4: Convert to binary based on gray level value
Step 5: If patches are present fill with holes
Step 6: Set a threshold value for opening area above some value.
Step 7: Use connected components method to determine the
number of objects present
Step 8: Draw bounding box
Step 9: Display output count and density
Step 10: End the program

14

4.3 VEHICLE COUNT :

Step1:Start the program.
Step2:Read the input background image of the empty road.
Step3:Read the new image with vehicles on road.
Step4:Convert
precision.

the images to grayscale format

using double

Step5:Find the width and height of the image.

Step6:Set threshold value =11
Step7:Find the difference between frames based on the threshold.
Step8:If frame diff > than 11 then assign that image to a variable
else 0 if no difference is found.
Step9: Increase the contrast of the output image using imadjust()
Step10:Find a graythreshhold value using graythresh() command.
Step11:Add Gaussian noise to the output difference.
Step12:Apply Weiner filter to filter the blobs
Step13:Convert to binary image
Step14:Fill holes to the blobs
Step15:Open all blobs having area greater than 5000
Step16:Count the number of cars using bwconncomp or bwlabel
Step17:Display the output image
Step18:Stop the program

15

CHAPTER 5
5. FLOWCHART:
5.1CELL COUNT:
START

GET INPUT IMAGE

CONVERTING COLOR TO GRAY IMAGE

USE CLEAR BORDER COMMAND

CHOOSE THRESHOLD VALUE
FILL BLOBS
APPLY MORPHOLOGICAL OPERATION
DISPLAY BLOBS
WHOS AREA >100
COUNT THE OBJECTS

CALCULATE THE DENSITY

STOP
Fig 5.1 Flow chart of the cell count
16

5.2 PEOPLE COUNT :

START

Getting the input

Changing color space from RGB to LAB

Binary to gray level value

If patches
are present
NO
Setting threshold value

Determining number of
objects present

Draw boundary box

Output display

End
17

YES

Fill with holes

Fig 5.2 Flow chart of the people count

5.3 VEHICLE COUNT:
START
READ THE BACKGROUND IMAGE
READ THE NEW IMAGE

CONVERT TO GRAYSCALE
FORMAT
DETERMINATION OF WIDTH AND
HEIGHT OF IMAGE
SET THRESHOLD VALUE=11
FINDING DIFFERENCE BETWEEN
FRAMES BASED ON THRESHOLD

ASSIGN 0 IF NO
DIFFERENCE IS
FOUND

NO

IF FRAME
DIFF>11

YES

ASSIGN A
VARIABLE TO
THAT IMAGE

INCREASING CONTRAST OF
OUTPUT IMAGE USING imadjust()

FINDING GRAYTHRESHOLD
VALUE USING graythresh()

ADDING GAUSSIAN NOISSE TO THE

OUTPUT DIFFERENCE
18

APPLYING WEINER FILTER

CONVERTING TO BINARY
IMAGE

FILLING HOLES

OPENING BLOBS OF AREA

GREATER THAN 5000

COUNTING THE NO. OF CARS

USING bwconncomp or bwlabel

DISPLAYING THE
OUTPUT IMAGE

END
Fig 5.3 Flow chart of the vehicle count

19

CHAPTER 6

6. MODULE DESCRIPTION:
6.1 CELL COUNT:
6.1.1 Image acquisition:
Imread() -Getting the input imagereads a grayscale or color
image from the file specified by the string filename. If the file is
not in the current folder, or in a folder on the MATLAB path,
specify the full pathname.

imshow(I,[low high])-

displays the grayscale image I,

specifying the display range for I in [low high]. The value low (and
any value less than low) displays as black; the value high (and any
value greater than high) displays as white. Values in between are
displayed as intermediate shades of gray, using the default number
of gray levels.
6.1.2Color to gray conversion:

graythresh(J(:,:,2)) -computes a global threshold (level) that

can be used to convert an intensity image to a binary image with
im2bw. level is a normalized intensity value that lies in the range
[0, 1].The graythresh function uses Otsu's method, which chooses
the threshold to minimize the intraclass variance of the black and
white pixels. Multidimensional arrays are converted automatically
to 2-D arrays using reshape. The graythresh function ignores any
nonzero imaginary part of I.
20

BW1=im2bw(J(:,:,2),L) -converts the grayscale image I to a

binary image. The output image BW replaces all pixels in the input
image with luminance greater than level with the value 1 (white)
and replaces all other pixels with the value 0 (black). Specify level
in the range [0,1]. This range is relative to the signal levels possible
for the image's class. Therefore, a level value of 0.5 is midway
between black and white, regardless of class.
6.1.3 Image enhancement:
imfill(BW1,'holes')-displays the binary image BW on the
screen and lets you define the region to fill by selecting points
interactively by using the mouse. To use this interactive syntax,
BW must be a 2-D image.

bwareaopen(bw2,100)removes from a binary image all

connected components (objects) that have fewer than P pixels,
producing another binary image, BW2. The default connectivity is
8 for two dimensions, 26 for three dimensions.
6.1.4 Connected components:

Cc = bwconncomp()- it used for calculating objects in image.

6.1.5 Density:

Objects per_unit_area=cc.NumObjects/(size(bw4,1)*size(bw4,2))Density is equal to connected components divided by size of image.

21

6.2 PEOPLE COUNT:

6.2.1 Getting the input image:
Imread() reads a grayscale or color image from the file
specified by the string filename. If the file is not in the current folder, or
in a folder on the MATLAB path, specify the full pathname.
6.2.2 Displaying the image:
imshow(I,[low high]) displays the grayscale image I,
specifying the display range for I in [low high]. The value low (and any
value less than low) displays as black; the value high (and any value
greater than high) displays as white. Values in between are displayed as
intermediate shades of gray, using the default number of gray levels.
6.2.3 Changing the color space:
makecform('srgb2lab') creates the color transformation
structure C that defines the color space conversion specified by type. To
perform the transformation, pass the color transformation structure as an
argument to the applycform function. In this case the RGB color space is
converted to LAB color space.

6.2.4 Finding gray thresh value:

graythresh(J(:,:,2)) computes a global threshold (level) that
can be used to convert an intensity image to a binary image with im2bw.
level is a normalized intensity value that lies in the range [0, 1]. The
graythresh function uses Otsu's method, which chooses the threshold to
minimize the intraclass variance of the black and white pixels.

22

Multidimensional arrays are converted automatically to 2-D arrays using

reshape. The graythresh function ignores any nonzero imaginary part of I.

6.2.5 Converting to binary image:

BW1=im2bw(J(:,:,2),L) converts the grayscale image I to a binary
image. The output image BW replaces all pixels in the input image with
luminance greater than level with the value 1 (white) and replaces all
other pixels with the value 0 (black). Specify level in the range [0,1]. This
range is relative to the signal levels possible for the image's class.
Therefore, a level value of 0.5 is midway between black and white,
regardless of class.

6.2.6 Filling patches with holes:

imfill(BW1,'holes') displays the binary image BW on the screen
and lets you define the region to fill by selecting points interactively by
using the mouse. To use this interactive syntax, BW must be a 2-D image.

6.2.7 Opening area greater than a value:

bwareaopen(bw2,1890) removes from a binary image all connected
components (objects) that have fewer than P pixels, producing another
binary image, BW2. The default connectivity is 8 for two dimensions, 26
for three dimensions.

23

6.2.8 Finding density:

bwconncomp(bw3) returns the connected components CC found in
BW3. The binary image BW3 can have any dimension. comp for finding
the density of people in image.
6.2.9 Bwlabel:
Labels connected components in 2-D binary image.
6.2.10 Measuring properties of image:
Regionprops(labeledImage,'all') measures a set of properties for
each connected component (object) in the binary image, BW. The image
BW is a logical array; it can have any dimension.
6.2.11 Drawing bounding box:
The imagesc function scales image data to the full range of the
current colormap and displays the image. imagesc(C) displays C as an
image. Each element of C corresponds to a rectangular area in the image.
The values of the elements of C are indices into the current colormap that
determine the color of each patch.

6.3 VEHICLE COUNT:

6.3.1 Getting the input image:
Imread() reads a grayscale or color image from the file specified by
the string filename. If the file is not in the current folder, or in a folder on
the MATLAB path, specify the full pathname.In this module we get two

24

input image one is the image with vehicles and the other is a background
image without vehicles in it.
6.3.2 Converting color image to gray:
Rgb2gray() converts both the images to gray images. Convert both
the images into gray level by using double precision.
6.3.3 Foreground detection:
Set threshold value=11, find the difference between the two images
by using abs() [absolute], abs can help find the absolute between the two
images. If the difference between the two images is greater than the
threshold value of 11 then those will be displayed as blobs at the output.
6.3.4 Morphological operation:
imageadjust(imadjust()) used to adjust the image intensity values to
the color map.
graythresh() used to set a suitable gray threshold value for the
output image.
Add Gaussian noise to the output image and filter it using wiener
filter.
Convert image to binary and fill holes if necessary. Open blobs of
area greater than 5000, this will help detect vehicles.
6.3.5 Counting and density calculation:
Count the no of blobs by using bwconncomp, the number of blobs
gives the number of vehicles present in the image. From the count the
density of traffic can be estimated

25

CHAPTER 7
7. SIMULATION RESULTS:
RESULT
7.1 RICE COUNT:
7.1.1 Input Image:
Image

Fig 7.1.1 Input image of the rice count

7.1.22 CONVERTING INTO BINARY IMAGE:
IMAGE

`
Fig 7.1.2 Binary image of the count

26

OUTPUT:
existing
Cc =
Connectivity: 8
ImageSize: [256 256]
NumObjects: 95
PixelIdxList: {1x95 cell}
Objects_per_unit_area =
0.0014

7.1.3 PADDING ZEROS:

Fig 7.1.3 Objects cleared at two side borders

OUTPUT:
padding
cc =
Connectivity: 8
ImageSize: [256 256]
NumObjects: 85
PixelIdxList: {1x85 cell}
objects_per_unit_area_2 =

27

0.0013

7.1.4 PROPOSED WAY OF CALCULATING DENSITY:

Fig 7.1.4 Objects cleared at four side borders

OUTPUT:
cc =
Connectivity: 8
ImageSize: [256 256]
NumObjects: 69
PixelIdxList: {1x69 cell}
objects_per_unit_area_3 =
0.0012

28

7.2 CELL COUNT:

7.2.1 INPUT IMAGE :

Fig 7.2.1 Input image of the cell count

7.2.2 CONVERTING INTO BINARY IMAGE:

Fig 7.2.2 Binary image of cell count

Output :
cellseg
objects_per_unit_area = 8.0601e-005
Connectivity: 8
ImageSize: [530 515]
NumObjects: 99
PixelIdxList: {1x99 cell}
29

7.2.3 Using proposed method:

Fig 7.2.3 output image of the cell count

OUTPUT:
cellseg
objects_per_unit_area =
2.0517e-004
cc =
Connectivity: 8
ImageSize: [530 515]
NumObjects: 56
PixelIdxList: {1x56 cell}

30

7.3 PEOPLE COUNTING:

7.3.1 Getting the input image:

Fig 7.3.1 Input image

7.3.2 RGB color converted LAB color :

Fig 7.3.2 Converted LAB imag

31

7.3.3 Equalise brightness to get skin area:

7.3.4 Converting to binary image:

32

7.3.5 Filling patches with holes:

7.3.6 Opening area greater than a value:

33

7.3.7 Final output:

Original with bounding boxes

50

100

150

200
50

100

150

OUTPUT:
cc =
Connectivity: 8
ImageSize: [225 300]
NumObjects: 7
PixelIdxList: {1x7 cell}
density = 1.0370e-004
numberOfPeople =7

34

200

250

300

7.4 VEHICLE COUNT :

7.4.1 Orignal Frame:

7.4.2 Converted Frame:

100
200
300
400
200

400

600

7.4.4 Foreground:

7.4.3 BACKGND Frame:

100
200
300
400

200

400

600

OUTPUT:

cc =
Connectivity: 8
ImageSize: [480 640]
NumObjects: 2
PixelIdxList: {[39316x1 double] [9899x1 double]}
6.5104e-006

35

CHAPTER 8
8. CONCLUSION:
In this project work, a simple way of counting objects in the images and
determining the correct value of density has been implemented. The number of
objects and the density values are determined and verified for cell, people and
vehicle counting applications. In cell count, the problem of counting many cells
automatically as well as determining the accurate value of density has been
solved. In people counting based on faces, even if too much skin is shown by a
person in the image the algorithm can detect faces correctly. This provides an
accurate count and density. Finally in vehicle counting, the algorithm used can
detect the number of vehicles accurately including two wheelers without use of
any hardware components.
In future, the same algorithm can be extended too many other
applications other than the three applications covered in this project. Also, the
present mat lab code used in cell count can be extended to find the correct count
even if cells tend to overlap in the image. In people counting, if the two faces
touch each other then inaccurate counts can be obtained. Hence methods to
segment such images can be developed. In vehicle count, the same method can
be extended to classify vehicles on the road such as car, truck, bike etc. Finally,
the same counting applications can be shown in a simpler manner using other
image processing software such as Python, ImageJ etc.
36

REFERENCES
Cell count:

[1] http://www.robots.ox.ac.uk/%7Evgg/research/counting/index.html.
[2] S.-Y. Cho, T. W. S. Chow, and C.-T.Leung.A neural-based crowd
estimation by hybrid global learningalgorithm. IEEE Transactions on Systems,
Man, and Cybernetics, Part B, 29(4):535541, 1999.
[3] C. Desai, D. Ramanan, and C. Fowlkes. Discriminative models for multiclass object layout. ICCV, 2009.
[4] X. Descombes, R. Minlos, and E. Zhizhina. Object extraction using a
stochastic birth-and-death dynamics incontinuum. Jthisnal of Mathematical
Imaging and Vision, 33(3):347359, 2009.
[5] L. Dong, V. Parameswaran, V. Ramesh, and I. Zoghlami. Fast crowd
segmentation using shape indexing.ICCV, pp. 18, 2007.
[6] M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman.
[7] T. Joachims, T. Finley, and C.-N. J. Yu. Cutting-plane training of structural
svms. Machine Learning, 77(1):2759, 2009.
[8] N. Ahuja and S. Todorovic.Extracting texels in 2.1d natural textures.ICCV,
pp. 18, 2007.
[9] S. An, P. Peursum, W. Liu, and S. Venkatesh. Efficient algorithms for
subwindow search in object detectionand localization. CVPR, pp. 264271,
2009.
[10] D. Anoraganingrum. Cell segmentation with median filter and
mathematical morphology operation.ImageAnalysis and Processing,
International Conference on, 0:1043, 1999.
[11] O. Barinova, V. Lempitsky, and P. Kohli. On the detection of multiple
object instances using Hough transforms.CVPR, 2010.

37

[12] J. L. Bentley. Programming pearls: Algorithm design techniques. Comm.

ACM, 27(9):865871, 1984.
[13] J. L. Bentley. Programming pearls: Perspective on performance. Comm.
ACM, 27(11):10871092, 1984.
[14] L. Breiman. Random forests. Machine Learning, 45(1):532, 2001.
[15] A. B. Chan, Z.-S. J. Liang, and N. Vasconcelos. Privacy preserving crowd
monitoring: Counting peoplewithout people models or tracking. CVPR, 2008.
http://pascallin.ecs.soton.ac.uk/challenges/VOC/voc2009/workshop/index.html.
The PASCAL Visual Object Classes Challenge 2009 (VOC2009) Results.

People count:
L. H. Liang, A. H. Zhou, G. Y. Xu, B. Zhang and L. H. Zhao, "A survey
of human face detection," Chinese Jthisnal of Computers, Vol. 25, No 5,
pp. 450-458, 2002.
1. P. Viola and M. Jones, "Rapid object detection using a boosted cacade of
simple features," Proceedings of IEEE Conference on Computer Vision
and Pattern Recognition, Vol. 1, pp. 511-518, 2001.
2. L. M. Luo, Z. Y. Peng and G. Y. Xu, "The feature of skin color," Jthisnal
of Software, Vol. 12, No 7, pp. 1032-1040, 2001.
3. R. L. Hsu, M. Abdel-Mottaleb and A. K. Jain, "Face detection in color
images," IEEE Transactions on Pattern Analysis and Machine
Inteligence, Vol. 24, pp. 696-706, 2002.
4. X. Tian, "Color space-based skin color subspace," Jthisnal of Xi'an
University of Science and Technology, Vol. 12, No 4, pp. 369-371,
2001.
5. X. S. Gao, X. Z. Zhang and C. Y. Ji, "Face detection in color images,"
Techniques of Automation and Applications, Vol. 24, No. 10, pp. 54-56,
2005.
6. Y. Freund, "Boosting a weak learning algorithm by majority,"
Information and Computation, Vol. 121, No 2, pp. 256-285, 1995.

38

7. P. Viola and M. Jones, "Robust real-time face detection," International

Jthisnal of Computer Vision, Vol. 57, No 2, pp. 137-154, 2001.
8. H. X. Liu and Y. C. Liu, "A face detection methord based on adaboost
algorithm," Jthisnal of Shanghai Jiaotong University, Vol. 42, No 7, pp.
1119-1123, 2008.

9. M. Yang and D. J. Kriegman, "Detecting Faces in Images A Survey",

IEEE Transactions On Pattern Analysis And Machine Intelligence,
Vol.24, No 1, pp.34-58, 2002.
G. Yang and T. S. Huang, "Human Face Detection in Complex
Background", Pattern Recognition, Vol.27, No 1, pp.53-63, 1994.
[CrossRef]
10.C. Kotropoulos and I. Pitas, "Rule-Based Face Detection in Frontal
Views", Proc. Int. Conf. Acoustics, Speech and Signal Processing, Vol.4,
pp.2537-2540, 1997.

11.Sanjay K. Singh, D.S. Chauhan, MayankVatsa, Richa Singh, "A Robust

Skin Color Based Face Detection Algorithm", Tamkang Jthisnal of
Science and Engineering, Vol.6, No 4, pp.227-234, 2003.
12.K. Sobottka and I. Pitas, "Face Localization and Feature Extraction Based
on Shape and Color Information", Proc. IEEE Int. Conf. Image
Processing, Vol.3, pp.483-486, Lausanne, 1996.

13.J. Miao, B. Yin, K. Wang, L. Shen, et X. Chen, "A Hierarchical

Multiscale and Multi-angle System for Human Face Detection in a
Complex Background Using Gravity-Center Template", Pattern
Recognition, Vol.32, No 7, pp.1237-1248, 1999.
[CrossRef]
[14] H. A. Rowley, S. Baluja, and T. Kanade, "Neural Network-Based Face
Detection", School of Computer Science, Carnegie Mellon University,
Pittsburg Research Center, 1998.

Vehicle count:
39

[1] R. Chellappa Vehicle detection and tracking

using acoustic and video sensors, 2004
[2] Sensor Line, Classax, Traffic Data Acquisition
System, Germany, www.sensorline.de
[3] S. Gupte, Detection and Classification of
Vehicles, University of Minnesota 2002.
[4] E. Atkociunas, "Image Processing in Road
Traffic Analysis", Vilnius University, 2005
[5] B. Coifman "A Real-Time Computer Vision
System for Vehicle Tracking and Traffic
Surveillance" University of California, 1998
[6] C. Zhang "Adaptive Background Learning for
Vehicle Detection and SpatioTemporalTracking", IEEE, 2003
[7] R. Cucchiara Vehicle Detection under Day and
Night Illumination" Proc. of ISCS-IIA, Special
session on vehicle traffic and surveillance, 1999
[8] Z.W. Kim "Fast Vehicle Detection with
Probabilistic Feature Grouping and its
Application to Vehicle Tracking", CiteSeer,
2001
[9] L. Di Stefano, Vehicle Detection and Tracking
Using the Block Matching Algorithm",
Department of Electronics, Computer Science
and Systems (DEIS) University of Bologna, ITALY

APPENDIX I
MATLAB CODING
40

CODE FOR CALCULATING NUMBER OF CELLS AND DENSITY IN

IMAGES

I = imread('dna.jpeg');%call image
bw0=imclearborder(I);
imshow(bw0),title('img 1');
bw1 = im2bw(bw0, graythresh(bw0));%apply threshold
figure;imshow(bw1),title('img 2');
bw2 = imfill(bw1,'holes');%fill hole in the image
figure;imshow(bw2),title('img 3');
bw3 = imopen(bw2, ones(5,5));%perform morphological ops in image
figure;imshow(bw3),title('img 4');
bw4 = bwareaopen(bw3, 100);%removes from a binary image all connected
comp of area less than 100 pixels
figure;imshow(bw4),title('img 5');
cc=bwconncomp(bw4)%count
objects_per_unit_area=cc.NumObjects/(size(bw4,1)*size(bw4,2))%Density

41

CODE FOR PEOPLE COUNT BASED ON FACES

%get the input image

I=imread('faces.jpg');
imshow(I),title('Image:1');
%change the color space
cform = makecform('srgb2lab');
J=applycform(I,cform);
figure;imshow(J),title('Image:2');
%equalise brightness to get skin area
K=J(:,:,2);% 2nd page of 3-d vector j
figure;imshow(K),title('Image:3');
L=graythresh(J(:,:,2));% find appropriate gray thresh value
BW1=im2bw(J(:,:,2),L);% convert to binary image based on threshold
figure;imshow(BW1),title('Image:4');
bw2=imfill(BW1,'holes');% fill patches with holes
figure;imshow(bw2)
bw3 = bwareaopen(bw2,1890); %opens area greater than 1890
cc=bwconncomp(bw3)% connected comp for finding the density of people in
image
density=cc.NumObjects / (size(bw3,1) * size(bw3,2))
figure;imshow(bw3)
labeledImage = bwlabel(bw3, 8);%same as connected components
figure;imshow(labeledImage)
blobMeasurements = regionprops(labeledImage,'all');%measure all properties of
the image
numberOfPeople = size(blobMeasurements, 1)% count the number of people
% draw bounding boxes
imagesc(I);
42

hold on;
title('Original with bounding boxes');
for k = 1 : numberOfPeople % Loop through all blobs.
% Find the mean of each blob.
% directly into regionprops.
thisBlobsBox = blobMeasurements(k).BoundingBox; % Get list of pixels in
current blob.
x1 = thisBlobsBox(1);%1st side
y1 = thisBlobsBox(2);%2nd side
x2 = x1 + thisBlobsBox(3);%3rd side
y2 = y1 + thisBlobsBox(4);%4th side
x = [x1 x2 x2 x1 x1];
y = [y1 y1 y2 y2 y1];
%subplot(3,4,2);
plot(x, y, 'LineWidth', 2);
end

43

CODE FOR VEHICLE COUNT AND DENSITY OF TRAFFIC

clc;
clear all;
MV = imread('cars1.png'); %To read image
MV1 = imread('backgnd.png');
A = double(rgb2gray(MV));%convert to gray
B= double(rgb2gray(MV1));%convert 2nd image to gray
[height width] = size(A); %image size?
h1 = figure(1);
%Foreground Detection
thresh=11;
fr_diff = abs(A-B);
for j = 1:width
for k = 1:height
if (fr_diff(k,j)>thresh)
fg(k,j) = A(k,j);
else
fg(k,j) = 0;
end
end
end
subplot(2,2,1) , imagesc(MV), title (['Orignal Frame']);
subplot(2,2,2) , imshow(mat2gray(A)), title ('converted Frame');
subplot(2,2,3) , imshow(mat2gray(B)), title ('BACKGND Frame ');
sd=imadjust(fg);% adjust the image intensity values to the color map
level=graythresh(sd);
m=imnoise(sd,'gaussian',0,0.025);% apply Gaussian noise
44

k=wiener2(m,[5,5]);%filtering using Weiner filter

bw=im2bw(k,level);
bw2=imfill(bw,'holes');
bw3 = bwareaopen(bw2,5000);
labeled = bwlabel(bw3,8);
cc=bwconncomp(bw3)
Densityoftraffic = cc.NumObjects/(size(bw3,1)*size(bw3,2));
blobMeasurements = regionprops(labeled,'all');
numberofcars = size(blobMeasurements, 1);
subplot(2,2,4) , imagesc(labeled), title (['Foreground']);
hold off;
disp(numberofcars);% display number of cars
disp(Densityoftraffic);%display number of vehicles

45

46