KARNATAK LAW SOCIETY’S

GOGTE INSTITUTE OF TECHNOLOGY

UDYAMBAG, BELAGAVI-590008
(An Autonomous Institution under Visvesvaraya Technological University, Belagavi)
(APPROVED BY AICTE, NEW DELHI)

Department of Electronics and Communication Engineering

Project Report on
Video-Based Vehicle Speed Measurement System
Submitted in partial fulfillment of the requirement for the award of the degree of
Bachelor of Engineering
In
Electronics and Communication Engineering

Submitted by
Prakash Gosbal 2GI17EC076
Saiprasad Wantmutte 2GI17EC105
Sampathkumar Jangamashetti 2GI17EC108
Shashank Gunjale 2GI17EC118

Guide
Dr. Mallikarjun H. Anandhalli
(Associate professor)

2020-2021

1
 KARNATAK LAW SOCIETY’S
GOGTE INSTITUTE OF TECHNOLOGY
UDYAMBAG, BELAGAVI-590008
(An Autonomous Institution under Visvesvaraya Technological University, Belagavi)
(APPROVED BY AICTE, NEW DELHI)

Department of Electronics And Communication Engineering

CERTIFICATE
Certified that the project entitled Video Based Vehicle Speed Measurement carried out
by Mr. Prakash Gosbal USN: 2GI17EC076, Mr. Saiprasad Wantmutte USN:
2GI17EC105, Mr. Sampathkumar Jangamashetti USN: 2GI17EC108, Mr. Shashank
Gunjale USN: 2GI17EC118, students of KLS Gogte Institute of Technology, Belagavi,
can be considered as a bonafide work for partial fulfillment for the award of Bachelor of
Engineering in Electronics And Communication Engineering of the Visvesvaraya
Technological University, Belagavi during the year 2020- 2021. It is certified that all
corrections/suggestions indicated have been incorporated in the report. The project report
has been approved as it satisfies the academic requirements prescribed for the said Degree.

Guide Co-Guide HOD Principal

Dr. Mallikarjun Anandhalli Prof. S.S.Saraf Prof. J.K.Kittur

Date:
Final Viva-Voce
Name of the examiners Date of Viva -voce Signature
1.
2.

2
 DECLARATION BY THE STUDENT(S)

We, Prakash Gosbal, Saiprasad Wantmutte, Sampathkumar Jangamashetti and Shashank

Gunjale hereby declare that the project report entitled Video Based Vehicle Speed
Measurement submitted by us to KLS Gogte Institute of Technology, Belagavi, in
partial fulfillment of the Degree of Bachelor of Engineering in Electronics and
Communication Engineering is a record of the project carried out under the guidance of
Dr. Mallikarjun Anandhalli. This report is for the academic purpose. We further declare
that the report has not been submitted and will not be submitted, either in part or full, to
any other institution and University for the award of any diploma or degree.

Name of the student USN Signature

1) Prakash Gosbal 2GI17EC076

2) Saiprasad Wantmutte 2GI17EC105

3) Sampathkumar Jangamashetti 2GI17EC108

4) Shashank Gunjale 2GI17EC118

Place: Belagavi
Date: 15/06/2021

i
 ACKNOWLEDGMENT

Any achievement does not depend solely on individual efforts but the guidance,
encouragement, and cooperation of intellectuals, elders, and friends. Some personalities, in
their capacities, have helped us in carrying out this project work. We would like to take
this opportunity to thank them all.
First and foremost, we would like to thank Dr. Mallikarjun Anandhalli, our
project guide. We thank him a lot for the guidance and cooperation which he provided
throughout the journey.
We would like to convey our heartfelt thanks to projecting coordinator, Prof. V. B.
Deshmukh, Head of the Department, Dr. Santosh. S. Saraf and the Principal J.K.Kittur,
for allowing us to work on this topic.
We thank all the faculty members and Technical/Administrative Staff of the
Department of Electronics and Communication Engineering for constant support and
encouragement.
We would also like to thank every one of those who have helped us in this
endeavor.
Last, but not least, we would like to thank our parents and friends who provided us
with valuable suggestions to improve our project.

Prakash Gosbal
Sampathkumar J
Saiprasad W
Shashank G

ii
 ABSTRACT

In our advanced transportation means detailed research of traffic data on highway

and roads is found to be as an important factor nowadays because of their cheap
prices and have perfectly developing abilities. Here we are using camera video
frames so that we can apply some new way of video processing and take the
expected data as output. Here in our project we are presenting a latest speed
detection approach of the vehicle i.e. by detection of motion. This method will also
be used to find speed of vehicle by deeply studying the frames of moving vehicle in
the region which we considered it as path which is of known dimensions. So this
method yields us better time outputs and works far well than old methods. Here in
our method we have three phases namely detection next tracking atlast speed
estimation. In our project we are using a mathematical model to find speed by
referring the video frames which vehicle passed taking it as input. When vehicle
enters path and gets detected and by transformation each frame is divided and
recorded with vehicle in it. Here we use technique of using intrusion line where we
can calculate the moving vehicle with least complexity. Next by mathematical
model which we use will get probability density function by which we can find
speed with more accuracy. By practical output we found our method has highly
reliable framework and accuracy unlike old methods.

iii
 Table of Contents
Table of Contents

Content Page No.

i. Declaration I
ii. Acknowledgement Ii
iii. Abstract Iii
iv. Table of contents Iv
v. List of Figures V
vi. List of Abbreviations V
vii. List of Symbols V

INTRODUCTION, OBJECTIVE AND 1

1.
METHODOLOGY
1.1 INTRODUCTION 01
1.2 OBJECTIVE 03
1.3 METHODOLOGY 03
2. LITERATURE REVIEW AND RESEARCH GAP 08
3. PRIOR ART SEARCH 15
PROBLEM DEFINITION, DESIGN THINKING
4.
APPROACH, FLOW CHART AND ALGORITHM
4.1 PROBLEM DEFINITION 28
4.2 DESIGN THINKING APPROACH 28
4.3 TOOLS AND TECHNOLOGIES 29

5. RESULTS AND DISCUSIONS 33

6. CONCLUSION AND FUTURE SCOPE 36

7. REFERENCE 38

8. APPENDIX 1 : SOURCE CODE 39

iv
 List of Figures
Figure No. Title ( Caption) Page No.
1 Basic block diagram for the project 04
2 Intrusion lines with line graphs 05
3 possible positions of intrusion lines 06
4 Real outcomes of the project 34

List of Abbreviations
Abbreviation Description
VTSM Video Type Speed Measurement
PDF Probability Density Function
CAMSHIFT Continuously Adaptive Mean shift
ITS Intelligent Transportation System
VSM Vehicle speed Management
GPS Global positioning Service
DHP Diagonal Hexadecimal Pattern

v
vi
CHAPTER 1: INTRODUCTION, OBJECTIVES AND
METHODOLOGY

1.1 INTRODUCTION:

By detailed research we found out that in roadway traffics, information is a key

element of today’s transport structure(7). In this paper we are describing a video
type speed measurement depending upon on a mathematical representation by
taking use of it we can find a drifting sample vector as data in it. Next we must
calculate speed of the vehicle which we will find by differentiating the path
(trajectories) of the detected vehicle in mathematical model which we are using in
this project(8). In this system, we make use of the intrusion line (line which
vehicle crosses) technique which is used to track the movement sample of the
input data with less complication. Later we will follow the mathematical model
where we can find out the vehicle’s probability density function relating to speed
which will increase momentum approximation, which gives us the output with
greater accuracy(6).
In this project we found that for finding the vehicle and measuring the speed, these
have a very keen importance in finding speed for all vehicles(11), but in these
systems, we are splitting them into intrusive sensors and non-intrusive sensors.
Firstly, In Intrusive type, which are commonly based on detecting inductive coil
which will be in a form of loop, which will be vastly applied, but they will be
having tougher fixation and heavy conservation, which gives us declining output,
and may also get destroyed. Next is Non-intrusive type will restrict these
problems, but they are basically more costly and which needs continuous
maintenance. As we know that dig cam are less costly and have capacity of creating
picture with higher accuracy, in video type models here in this project we will be
using lower cost camera and items for measuring the speed of non intrusive
vehicles. In our project we are going to explain the non-intrusive type of video
models for detecting vehicle and measuring its speed(7). Our aim for this project is
detecting momentum of the vehicle with high exactness by analyzing video frames
which we have taken as input to track the automobile and find the speed(4) .

Let us make an assumption that vehicle moving in path having speed which we take
it as constant, then input video can be captured by a fixed high-up camera, which is
positioned so that the camera facing down towards the vehicle path where it will
capture vehicle passing as input(15). The common techniques which we can use
are foreground /background segmentation, where we can track the moving vehicle
in order to detect its speed in continuous frame by frame(3). These frames we can
study by applying computer vision technique which is used to find out the

1
driftness 0f the vehicle passing the zone. In a background model, we create road
visually and then the camera vibration will be there which is balanced to decrease
the noise and accordingly improve the accuracy of speed. We need to track the
every single pixels of vehicle passing the area of detection and keep away the fake
things like shadows, noise etc. Since the video which we captured as input comprise
of consecutive frames in constant time interval(1). Therefore, this analytical model
gives us the output with a probability density function (PDF) which measures
momentum of a moving automobile(13).
Our important aspect in project was always for finding and tracking all the
properties needed those all which constitute vehicles passing the path and to reach
higher precised outcomes by finding out the vehicle properties on the likeway path
for different measurements. The LK (Lucas Kannade) algorithm rule, Horn
Schunck technique and Farneback quadratic extension algorithm, these are the
commonly used alogrithms which we can use to achieve the optical flow . Mean
Shift and Camshift (Continuously Adaptive Mean-Shift) technique(5) are also
some of the important properties which we can use in detecting the vehicles where
the dimensions may deflect in successive frames ; otherwise, we may find it
difficult to get an absolute geometrically correct output values(2).
There are also processes which we are following by above any algorithm where
we can find the driftness of the vehicle passing on the path , for eg.., by detecting
license plate area or place. For tracking number plate we need minimum two
images containing the passing vehicle. From using these above algorithms we will
be required with specific camera to be placed and images with high quality
resolution and pixel by which we can detect the license plate. In this model we will
examine samples taken in time interval by the camera, the positioning of intrusion
lines and we can atlast find the vehicle’s speed. In this paper we conclude by using
the rate at which frames are caught by camera by positioning of all input intrusion
lines and a drifting sample input data which helps us to give output in form of pdf
(probability density function) type of the vehicles momentum. These outcomes we
are discussing and evaluating are dependent upon congestion region rules and
orders passed by the recommended national lane authority for variety of
automobiles.
In our project we are discussing about the ability that how to cope with the samples
taken in time interval by the camera and the vehicle’s located and speed
estimation, which makes us capable of how to measure the driftness in a standard
valuable aspect. The model gives us many more intrusion lines, accompanied by
their specific distance, where it helps us to identify those motion pattern of a
vehicle crossing the lines. Further, in our discussed project we are going to evaluate
the PDF of vehicle fastness by utilising the found input motion design. The
important aspects we learned firstly is, Speaking about the parameters deflecting

2
the speed values by analyzing video type models and, lastly expanding a
extensive mathematical procedure for calculating probability density function
which helps for calculating the vehicle’s speed in the footage captured.
This model mainly aims to put forward a creative type detection of the speed in
which we can cover larger surface and place or we can install in multi-lane
highways with simple installing means and it also takes least conservations and
cost friendly, when we differentiate them by active type of sensors and by using
this concept or project we can calculate the speed of a motor devices with more
accuracy which is also using only few of input parameters taken into consideration
and with also by giving out least measurable complications.

1.2 OBJECTIVE:
Our work is having a motivation to introduce a new type of video based speed
detection model which is going to cover a vast distance or on a multi-lane highway
by least advanced installation and also least maintenance costs when compared with
radar type active sensors. Likely other motivation is, the proposed project is used
for finding the vehicle’s driftness almost accurately with using some of many
parameters and with a least complex outputs.
The main Objectives of the paper are listed below:
 To outline an instinctive and multifaceted system which directs to certain
issues like vehicle’s overspeeding without any use of man wealth and assets.
 Speaking about the parameters which affects speed detection and finding
values in our motion based vehicle speed detection model systematically.
 To find out the speed of vehicle passing with crossing speed limits and find
the license number plate of the vehicle using image method.
 Expanding the use of the mathematical model by which we can attain
probability density function of automobile driftness in the frames captured in
video.
 To setup the connections between different modes and models or systems for
exchanging the input and output data in system.

1.3 METHODOLOGY:
In this model we are going to consider some of the parameter which are helping
their part to uncertainty in the speed measurement system. The detection of vehicle
depends upon the scenes by using the intrusion lines taken from video source and
knowing about the positioning it has in ruling life world and vehicle passing in the
region as shown in the figure. Further vehicle driftness is taken as fixed value while
vehicle is clearing the path because the scenario used has very least measurements
that too in meters. The model which is introduced here will give off the probability
density function of vehicle speed which we find using moving pattern input frames,

3
using these steps is discussed later in model. By using lines and also keeping those
moving pattern inputs will yield us incredible output in both that’s memory
effectiveness and calculations of complex nature.

Fig (1) A detection scene containing several intrusion lines (lm+1 , where m ∈ {0, . . . , M − 1})
and a passing vehicle observed by a camera facing the road.

1.3.1 Speed estimation model:

Line type finding which technique is applied for finding out if the vehicle passes
the line virtually created and will enter the region of frames. In this project for
detecting the videos findings, these vehicle which are passing is detected at some
different places because of the temporal sampling type rate. The time passing of
sampling between the found different locations is same and equal to camera
sampling driftness. So only, the intrusion line is unable to detect the line not
accurately but by some distance detected (m). So the distance detected is directly
proportional to sampling time of the camera T(s) and the driftness which is written
in (m/s).
i.e. Δ= Tv ------------------ (1)
the above equation relates that the distance detected is proportional to the driftness
of the vehicle and the rate of frames which are inversely to the camera. Then the
starting point (initial position) x1(m) when vehicle enters the first intrusion data is
taken as random. So in some cases of more intrusion lines which are almost
accurate in finding out only when automobiles can pass the intrusion line and
distance which is considered with each other successive intrusion lines. So in our
system, a model is introduced to find out probability density function of a vehicle
crossing speed which is depended upon movement type of input. Therefore, we call
M as intrusion lines so M≥ 2 must be placed on the input video frames of receiver

4
end of the camera located constant such that distance between them in the running
world on the road is to be known. According the vehicle passing is detected as
soon as it passes the every specific line at definite frame value.

(a)

(b)

(c)
Fig. 2. (a) The intrusion lines lm , (b) The hypothetical vehicle’s positions xm within the
maximum detection distance O for a hypothetical speed v, (c) One illustration of
vehicle’s positions and the associated nm s for a hypothetical speed v.

The important inputs of our project is detected by moving vector pattern i.e. n= [n 0
. . . n M-1] € N next the distance input of the intrusion lines
d=[d 0, . …., dM-1] € R , ------------(2)
where we call it as , n m = f m +1 −f 1 and d m = l m +1 −l 1 where m ∈ { 0 , . .
. , M −1 } , M is given as number of intrusion lines, fm+1 is the input frames of the
video by which we can detect and lm+1 is also the location of the lines of intrusion
in the ruling world systems . As we discussed earlier the related position of the lines
of the distance vector d and the frame rate of the video captured 1/T, a vehicle when
is detected at any one of the position of the intrusion line. So only the initial point
distance ϭ from 1st initial point is not known and it is not more than maximum
distance to be detected where 0≤ϭ≤1. So only vehicle placement xm can be
detected using the initial point computational i.e. x1, frame rate, input vector of
distance and speed of the vehicle.

5
 (a)

(b)

(c)

FIG(3) One example of possible positions based on: (a) the distance vector of the
intrusion lines d, (b) the movement pattern vector n, and (c) the overlap intervals for
(a) and (b) for a hypothetical speed v.

. So further we will get 2 types of the placements which helps the position to
demonstrate vehicle with speed v which is taken by the camera with the frames rate
i.e. 1/T which is covered along the way travelled.
 First some of the positions are depending upon the distance vector and the
intrusion lines. These placements show the detection of the distance where
vehicle will be travelling with some speed which is considered to nullify the
entire real life distance vector by use of intrusion lines (as shown in figure).
 Next some of the position is detected using moving pattern of the vehicle n.
Then position shows us detection of the vehicle passing some distance which
satisfy the moving pattern of the vehicle (as shown in figure).
So these are 2 needed conditions to detect the above intersection for all distance for
every intrusion line detected. So next we will be discussing about implementation
about speed measurement system.

1.3.2 IMPLEMENTATION OF SPEED ESTIMATION SYSTEM:

 As discussed earlier, the movement of the vector input n is a group of frames
of the video difference i.e. the total frames which are between findings at
consecutive intrusion lines. So only detective system are required for every
single intrusion line to determine the vehicle entering the some of the
distance which is to be detected. Firstly, the important features of the
highway lane respective to the area which is taken as frame area which
receives video which are needed to be taken and extracted. Then these points
are detected in successive frames and region of motion of vehicle is detected.
 Motion Detector :

6
The important output of movement algorithm is used to find speed or
displacement of the vehicle to identify the movement of vehicle in every
single frame. So here in this paper we use movement detection method
which is based upon polynomial expansion algorithm. So this polynomial
algorithm uses an optical type which is a descriptive system which almost
gives image pixel as function, and by using this technique we use the image
as a continuous variable, and this small speed measurement of the picture is
calculated by equalizing the variables of the polynomial expansions at one
point.
By using the method we can use movement estimation component by using
which we can find the frames of the picture when the vehicle will be passing
the created intrusion lines. So only the speed is used to find out the
importance of the movement by utilising the distance to find out the vehicle
passing the intrusion lines where the noise is ignored and even unwanted
disturbances. For all the discussed types, vehicle which passes the zone
which is marked for finding out after every single intrusion line is used for
at greater possible driftness ( v max ) which might be 40 m/s. So only, the
model we use can detect the automobile which is moving with the given less
speed, i.e. the maximum detection distance is taken as region of interest for
the driftness which is less than 40m/s. So the pdf of the automobile speed is
then determined based on the detected movement pattern vector n.

7
CHAPTER 2 : LITERATURE REVIEW AND RESEARCH
GAP

1. Vehicle speed measurement (VSM) is depended upon video pictures which

are used to represent the development of speed measuring in the upmost
intelligent transportation systems. Thus this project is implementing a
different type of driftness calculating method of the automobiles, which
contains the improvement of three frame difference algorithm and the
proposed gray constraints optical flow algorithm. By then improving three
frame algorithms, the moving vehicle type of counter can be found out
much easier way. By proposed gray model constraints optical flow, the
optical flow value which is driftness of the vehicle which is to be found in
images. Further the velocity of the automobile is finded out by the optical
flow of the vehicle contour and the pixels of image and road wideness
ratio. This method will yield a better optical flow by reducing the influence
of difference of the lights and error in form of distraction shadows. It will
reduce the results because to calculate the targets optical flow value. So
this proposal type of approach is having a average type of estimation
vehicles speed.

2. In this work author shares information about an highly efficient and strong
powerful type of system which is used to find out the limits of the speed,
where it will give us higher accuracy and at least time for processing.
There are best methods to utilize these in every step i.e.
 Speed limit digit detection: the method which is introduced to step
in as image processing type for the use of detection of signs using
the speed limits which is based on the working of the haar like
accessibility.
 Speed limit digit recognition: here we have checked the multiple
classifiers to divide it to a higher efficiency and most powerful type
of model systems which will help us to find out and detect the speed
limits signs. Then further we can use k-NN algorithm to find the
speed of the moving vehicle is said by author in his work.

3. There are some systems which are used by parking lot access control
which are developed in vehicle management by using tracking and number
plate detection. These systems author is using camera to detect the vehicle
by detecting vehicle by using license plates depending upon optical
character detection technique. In this paper author presents a new type of
system to detect and identify the Brazilian license plate vehicles in which
vehicle users have permission to enter the location. For this, techniques of
Digital Image Processing were used, such as Hough Transform,
Morphology, Threshold and Canny Edge Detector to extract characters, as

8
 well as Least Squares, Least Mean Squares, Extreme Learning Machine,
and Neural Network Multilayer Perceptron to identify the numbers and
letters. So author states that system has tested with 700 videos with a
resolution of 640*480 pixels, which grants the access only when number
plate is registered with highest successful rate of 98.5% on tested data
bases. The movement detection step is connected to system which is faster
and more precised in the real world. So the author concluded that this
model is an important tool which has high standard which can be used
nowadays commercially.

4. Automatic video analysis from urban surveillance camera is the quick

emerging field which is dependent upon computer vision technique. So
here author presents a comprehensive result of the state of art computer
vision used for traffic video which has an outlook to research directions.
So this field is increased the importance in intelligent transport systems
(ITS). So the lower hardware economic and thus the increase deployment
of the high quality camera which have given idea to open wide
applications for the analysis of the videos.
There are many objectives and problems in the paper said by authors like
congestion, traffic rules which are been violated and vehicle interactions
can be targeted by the use of the cameras which are really installed by
human operators. Systems which are used for detection and classification
of these vehicles on the highways which have been installed by many
surveillance camera technique like background detection and motion
tracking of the vehicle for some time. The urban domain is more difficult
with respect to traffic density, different camera frames and camera angles
deflections which lead to higher occlusions and the methods for object
categorization and 3D Modeling has inspired higher techniques to meet the
difficulties. Further author states that under challenging weather condition
like fog, rain will be helpful but it helps future work should be made
towards robust detectors and for all road these type of realistic conditions
this type of evaluation can be done is stated by the authors.

5. For evaluating the effectiveness of using position of the cell along with
the GPS sensors type based upon collecting by experiments on energy
utilization and finding location more precisely for some sort of data
acquisition modes where only GPS positioning i.e. by only cellular
positioning and both methods are taken into action. So our model is
simulated to find the average velocity calculation for extraction for
locating data with hybrid data acquisition model. Our proposed system is
made to be utilized for finding average velocity determination which is
essential variable in traffic monitoring models. Here we assume that the
average speed acquired from the gps positioning is taken as ground truth to
know the activity performance of the model. By assessing the feasibility of

9
 the cell position where the GPS sensors which are not considered as
available for location approximation.
So in this paper author suggest that a hybrid acquisition model is applied
for cellular positioning where GPS sensors technique which helps to obtain
the raw location data where gps sensor is not applicable or battery of
particular smart phone is too low for location detection. For modeling the
traffic flow and gathering the traffic congestion information are the two
difficulty and challenges nowadays in smart transportation system. Most of
the traffic models and detection of velocity are methods which are been
proposed in this project by the author for gathering the data from GPS
equipped smart phones and earn back the model which is depended upon
GPS sampling techniques. Then too approaches made by authors are tend
to failure because of the insufficient vehicle input data and undetermined
dynamics of the flow of vehicles. Next by using the gps sensors might
make way for battery draining and it will reduce the overall performance
the system. So in this paper work author says that to propose a new battery
life friendly data acquisition model which is used to get the new input data.
So the author says that this model has yielded them output of 88%
accuracy to know the location of the user and even the consumption of the
battery is reduced and gave the expected output.

6. In this paper author has addressed about the problem which is faced by
them to measure the speed of the vehicle just by seeing the video which is
captured by camera in an urban cities. So they proposed this type of
system based on which we can find the selection and tracking of the
distinctive features which is been placed in every single vehicle plate
region. So this model was tested on the almost of 5 hours of the video
based hd quality views and checked on around 8000 vehicles in three
multilane highways, which is treated as ground truth and taken as highly
accurate which is depended upon loop detectors. In future work, they
decided to cross check that by estimating the license plates on reality
which might yield high results. Authors also said that they aimed to apply
OCR on the license plate which are detected so that they can create a
traffic speed control systems with highly equipped surveillance tools; for
ex. To know about the flow of the traffic, to identify the vehicles which are
stolen etc. One more important topic for future is that to implement the
compact stage which gives permission for local processing, including
optimizations such as parallel processing on GPUs, thus reducing
communication bandwidth requirements.

7. In this project author have discussed the issues of vehicle to find out the
speed depending upon videos which is taken by camera is fixed in urban
places. Here there has been introduced a model which is dependent on

10
 selecting and detecting and tracking of the instinctive characteristics which
is situated in every single vehicle license plate region. So in this system
they have almost tested with every single type of the vehicles that is bike,
cars and heavy vehicles too with full high quality, with unlimited data sets
created by us and detected on multiple road lanes. , which they have stated
in our paper is all real values which we have taken with higher accuracy
system which is based on many type of inductors loop detectors and these
all are truth which is labeled to license plate regions. So in this project
authors have detected license plate region which yielded them the better
results rather than taking the vehicle.
The project is depended upon speed detection by seeing videos which gives
us the probability density function of automobile which crosses with some
speed based on its moving way or path input. So in this work they have
proposed a new type which has been used by arranging some of the intrusion
lines and taking it as data input frames which they have captured
simultaneously by using the camera which we have fixed on upper hand of
the road with least minimal installation and by using Smartphone high pixel
camera with each frame rate of 50fps and 40fps respectively. So when they
have taken the project details and results, the error rate of the vehicle speed
is at 1% and 2% respectively, for vehicle which is moving with speed of 60-
80km/hr. So that they can reduce error rate when can be done by increased
number of intrusion lines which says that gave more confidence in this
project. The detection region which we constructed with different intrusion
lines for each and every input we made as line. The vehicle passes the
intrusion lines using the model analytically. This project when computed by
using the pdf of the vehicle speed which is used by movement of the vector
input, intrusion lines and the sampling rate and time of the camera. So the
proposed model which they used is so compatible to older methods and they
can improve the deflection and precision because all the ideas which
measure as displacement for some period of the time based on the frames
too. By using this proposed model they found the speed of the vehicle by
using some video frames.

8. So here in this project authors states that this system is focused on

employing the multiple intrusion lines even though the parameters are
evaluated under their own circumstances which made this as one of the
complications. The project is depended upon speed detection by seeing
videos which gives us the probability density function of automobile which
crosses with some speed based on its moving way or path input. So in this
work they have proposed a new type which has been used by arranging some
of the intrusion lines and taking it as data input frames which they have
captured simultaneously by using the camera which we have fixed on upper
hand of the road with least minimal installation and by using Smartphone

11
 high pixel camera with each frame rate of 50fps and 40fps respectively. So
when they have taken the project details and results, the error rate of the
vehicle speed is at 1% and 2% respectively, for vehicle which is moving
with speed of 60-80km/hr. So that they can reduce error rate when can be
done by increased number of intrusion lines which says that gave more
confidence in this project.

9. The proposed methodology in this project is discussed by the author who

involves the vehicle speed estimation which is taken as using the image
processing technique. Therefore the vehicle speed estimation is also a
technique which is carried out by using the different approaches that is
moving for multi detection of the vehicle, tracking the detected vehicle
which results in the higher accuracy. So this is the experiment which is
calculated out on 3 different video samples format. The multiple vehicles
which are detected by the gradient edge detection and the normal cross
correlation algorithm is used by the vehicles are tracked using 2 frame
difference algorithm. So finally the vehicle speed is estimated by use of
features for tracking the vehicle’s distance travelled by using the frame
rate. The author shares that the performance of the system which is
introduced is calculated by using four higher precision measures i.e.
accuracy, precision, false positive rate and error.

10. In this paper author says that he used a new type of algorithm for the
detection of the of the individual vehicle speed based upon two
consecutive images which are to be captured from camera system which is
used for traffic safety. These are the principles first where both images are
transformed from image plane to the 3D world coordinates which is
depended upon calibrated camera parameters, and secondly the difference
between the two images which are transformed is determined which will
be resulting in the background being removed and vehicles in the two
images mapped onto one image. Atlast a block feature of the vehicle
closest to the ground is made a match to estimate vehicle travel distance
and driftness. So experimental results show that the proposed method
exhibits good and consistent performance. When it is compared with speed
measurement obtained from other speed radars which is averaged
estimation errors are 3.27% and8.51% for day time which was detected
respectively, which are so better than previously published results. The
proposed algorithm can be easily extended to work on sequences of the
images.

11. In this paper author states a technique to estimate vehicles velocity using
the images which is rectified that represents a top view of the highway. To
rectify the image sequence captured by uncalibrated cameras, so authors
says that this method automatically estimates two vanishing points using
the lines from the image plane. So this approach requires two lengths
12
 which is known on the ground plane and can be applied on the highways
that are nearly almost straight near the camera is kept for surveillance.
Once the background image is rectified type which is possible to locate the
stripes and the boundaries of the lanes at the highways. So this process
may be used to count the number of the vehicles which is used to estimate
the respective velocities and the mean velocity which is also associated to
each of the highways which are identified first only is shared by authors in
this paperwork.

12. The vehicle tracking and speed measurement system which is proposed to
find out the speed of the vehicles by considering some of the parameters.
The authors shares that there is new feature added that is DHP is proposed
for use in the both background and foreground images. DHP detects the
spatial texture information based diagonal derivatives. Edges and corners of
each object extracted using DHP act as feature for tracking. Object detection
and tracking is a vital role and active research area in computer vision
technology. This proposed tracking of vehicle and measurement of speed
system which can find out speed parameters of the vehicles. So speed
parameters which are used to take judgment to find the speed parameters of
the accident which causes at low cost. So here author says that main
objective is to develop an algorithm which will help to detect foreground,
track specified objects and even calculate the parameters of the speed of
objects. Even identifying stationary background from moving the objects in
a video is an important task. To achieve superior foreground detection
quality across some of the scenarios which forms the dynamic background
for object tracking which will be further used in as diagonal hexadecimal
pattern(DHP). There are many more measures to be used for the
performance of the proposed system. So the author says that this type of
model will yield a higher result for tracking and estimation.

13. Here the author says that identifying moving objects in a video sequence
is a fundamental and important task in as much as computer-vision
applications. Automatic video analysis from urban surveillance camera is
the quick emerging field which is dependent upon computer vision
technique. So here author presents a comprehensive result of the state of
art computer vision used for traffic video which has an outlook to research
directions. So this field is increased the importance in intelligent transport
systems (ITS). So the lower hardware economic and thus the increase
deployment of the high quality camera which have given idea to open wide
applications for the analysis of the videos.
There are many objectives and problems in the paper said by authors like
congestion, traffic rules which are been violated and vehicle interactions can be
targeted by the use of the cameras which are really installed by human operators.So
the author has proposed algorithm which does not perform well when there are
multiple moving objects close to each other. They have built their algorithm by

13
building multiple ellipses to identify moving objects and using the background
distribution to identify the background area among them. They also included
additional sequences taken under different conditions, as well as additional ground-
truth frames, to test how well these enhancements perform in real traffic sequences.

14. Background subtraction is an important step for analysis of videos which is

confessed by the authors, where it is used to find out the objects of interest
for further processing. So here this algorithm is often included of
background model and a common regular type of pattern. This background
model will help to determine a per pixel measurement of the pixel if it
belongs to the background or foreground, while the regular model brings in
information from the adjacent pixels. Ground-truth experiments with urban
traffic sequences show that our proposed algorithm significantly improves
upon results using only Kalman filter or frame-differencing, and outperforms
other techniques based on mixture of Gaussians, median filter, and
approximated median filter the proposed algorithm does not perform well
when there are multiple moving objects close to each other. So the author
describes that whether its background model or foreground the output of this
paper will yield higher better results comparing to the older method used.

15. Here the author has the rapid recent advancements in the computational
ability of the everyday computers and new technologies have made the work
possible to widely apply deep learning method to analysis of traffic
surveillance videos. In this paper authors said that traffic flow prediction,
anomaly detection, vehicle reidentification and vehicle tracking are some of
the basic components in analysis of traffic management. So among the many
applications traffic detection, vehicle detection and speed measuring all are
mostly important used by authors in the project. In the 2018 NVIDIA AI
City Challenge, they have combined modern deep learning models with
classic computer vision approaches to propose an efficient way to predict
vehicle speed. In this paper, authors have introduced some state-of-the-art
approaches in vehicle speed estimation, vehicle detection, and object
tracking, as well as the solution which is taken as in this project.

16. So in order to find the speed of the moving vehicle by using of the side
views of the camera images, by hat authors have used velocity vectors of a
sufficient number of the reference points which is identified on the vehicle
by using the frame images. Authors said that it has 2 steps: firstly, a
sufficient number of points from the vehicle are selected, and these points
must be accurately tracked on at least two successive video frames. In the
second step, by using the displacement vectors of the tracked points and
passed time, the velocity vectors of those points are computed. So this type
of computed velocity is defined by authors as in video image coordination
system and displacement vectors which are measured in pixels units.
So the author says the magnitude of the vectors in image space which should
be transformed to the object space where we can find absolute values of
14
 these magnitudes. These transformations requires an object space
information in a very mathematical sense that is achieved by means of
calibration and orientation parameters of the images which are taken in video
frames is explained by authors in this paper.

CHAPTER 3: PRIOR ART SEARCH (PATENTS AND

COPYRIGHTS)
PATENT 1 :

TITLE : VIDEO-BASED VEHICLE SPEED ESTMLATION FROM MOTION

VECTORS VIDEO STREAMS

APPLICATION NUMBER: 13/600,742

PATENT NO : US 9,582,722 B2
INVENTORS: Edgar A Bernal, Webster, NY (US)
WHENCHUNG WU , WEBSTER NY(US)
ORHAN BULAN, GREECE , NY(US)

APPLICANTS: Robert P. Loce, Webster, NY (US)

DATE OF PATENT GRANTED: Feb. 28, 2017
COUNTRY: UNITED STATES PATENT
ABSTRACT: It has been automated with less difference is also video based speed
finding of the vehicle which helps in operating within the streams of the video for
finding out all the screen sequences for identifying and eliminating all the non clear
violators and find out all the listed rule breakers by which a multilayer speed
enforcement systems are used, in which we can find deeper layers which will
provide us a more enhanced accuracy on some of the selected vehicles with some
speed values. Here video motion type clusters are used with vehicles are identified
and tracked and across multiple frames which is taken from video which is taken by
the camera. So from the various types of the vehicle position detected and by which
speed can also be estimated which is fully based upon geometric data which is
available of some vehicles but unlike some characteristics like end points of the
license plate characters and when the points or characters of the license plate are
totally intact. So here auto estimated speed data is added which helps the video
captured as a type of data i.e. metadata, and video segments of vehicle speed values
are stored and sent for the automated or manual reviewing.

CLAIMS:
1. A computer-implemented type for detecting the speed of the vehicle from
motion vector information which includes:
 receiving video data of a moving vehicle from a camera;

15
  extracting motion vectors in the data which can be extract;
 identifying and analyzing at least one motion vector cluster representing the
moving vehicle; and
 tracking at least one feature of the at least one motion vector cluster across a
plurality of video frames; and
 estimating frame-to-frame speed and mean speed of the vehicle which is
based at least in part on spatio-temporal coordinates of the at least one
tracked feature with the aid of calibration data.

2. The procedure which was used at claim 1, wherein the motion vectors are used
as one of the input type of motion vectors used in the method of compressing
the input data set.
3. The procedure which was used at claim 2, for estimating each and every frames
driftness and finding the average speed of the vehicle which is done by combining
the vector lengths which are measured.

4. The procedure of 3, which we have used the mixed type of vector inputs of
different lengths is of an average type and is an order of statistic type.

5. The procedure which is meant for claim 1, by using where the motion vector
form of the cluster is regular form to delete the input vectors which are considered
for never to correspond for expected vehicle trajectory.

6. The procedure which was used for claim 1, where received dataset in form of
videos is fully compressed form of data video set and further we can identify the
motion vectors which contains motion input data which we have taken it from the
video data set which is now compressed.

7. The procedure which was used for claim 1, where we have got the video input
data as dataset which is fully comprised of fully uncompressed data with a live
stream, and wherein received data and motion vectors are to identified which are
usually created during the compression of the video frames.

9. The method with respect to claim 1, where we can even find the range of
estimation for mean speed of the vehicle which will fall outside the speed range of
the vehicle which is to be taken as predetermined in sense of transmitting a video
frame by frame, by which vehicle can be citied and sent for speed detection.

10. The procedure which was used for claim 1, which is further comprised of
transmitting one or many more types of segments of compressed video input form
of data which is having estimated speed information and it’s even embedded with
to a remote position for accurately noticing of the vehicle speed, where we have got
the one or more transmitted datasets segments which is highly comprised of the
compressed video frames in it and one of or more than one moving vehicles are
detected for speeding values.

16
11. The procedure which was used for claim 1, which is further comprised, when if
it is determined for average speed of the automobile will cross a value of decided
speed called as threshold for vehicles, which helps to perform speed estimation
which is also called to be on a fully decompressed type of the data used in the
videos.

12. The procedure which was used for claim 1, where the camera is attached to a
top of the stationary and constant structure.

13. The procedure which was used for claim 1, where the camera is fixed on the
vehicle top for capturing the video.

14. The type of processor which was used to execute in the processor with some
instructions for performing the same method which was used in claim 1, stating that
instruction being kept preserved are computer read able only type.

PATENT 2:

TITLE: Vehicle speed detection

APPLICATION NUMBER: 14/616,115
PATENT PUBLICATION NUMBER : US
USUS20160232410A1S20160232410A12 20160232410A1
0160232410A
INVENTORS: Michael F. Kelly, Montreal West (CA): David McMordie,
Montreal (CA)
DATE OF PATENT PUBLISHED : Aug. 11, 2016
COUNTRY: UNITED STATES PATENT
ABSTRACT:
For the improved system and methods which are used to determine the speed of the
vehicle on the multilane highway which is precisely used of sequence of the real
time of the images. So from the various types of the vehicle position detected and
by which speed can also be estimated which is fully based upon geometric data
which is available of some vehicles but unlike some characteristics like end points
of the license plate characters and when the points or characters of the license plate
are totally intact. So the bodies of the invention is going to be facilitate when
position of the vehicle and also the driftness which is calculated without any help of
any manual or by calculating of the position of the camera which is located in space
when compared with respect to the road area.

17
CLAIMS:
1. There are some of the methods which are used for detecting the speed of the
motor vehicles, these are the steps which are comprised of:
 providing a basic type which is comprised of one type of the
camera’s videos where they can be used for capturing a number of
series of successive where its separated by time which images will be
having a features of physical type.
 Which is used for knowing the location of features within which each
of the images are time separated;
 based on the known geometric parameter of the feature and a
geometry of the camera relative to the vehicle, estimating, for each of
the time-separated images, a real-world spatial coordinate position of
the feature; and
 based on the estimated real-world spatial coordinate positions of the
feature in the time-separated images and a capture time of each of the
time-separated images, estimating a speed of the moving vehicle.

2. The procedure which is applied at claim 1 is that where the features which
contain at least two identifiable features point.

3. The procedure which is applied at claim 1, where the features which are
considered as top and bottom of at least a single character on the license plate
region of the vehicle.

4. The procedure which is applied at claim 1 where the steps are fully comprised of
the basic height of the character which is to look up for the characters on the
number plate.

5. The procedure which is applied at claim 1, where it is fully comprised of the

steps which will help to perform the normalization of transforming by the image
prior for determining the steps.

6. The procedure which is applied at claim 1, which is fully having the distance
from the camera to which the character used for pairs of ray type which is totally
depended upon distance between the characters of the points.

7. The method of applying claim 6, later on which is comprised of the step which is
used for creating their own trajectory which is shown as the real world coordinate
points of the real imagining world.

8. The procedure which is applied at claim 1, where the driftness of the vehicle is
approximated by adding the rule of linear or fitting of the curve algorithm to all the
trajectories.

18
9. The method which is applied at claim 7 is to make the trajectories which are
made up at each and every feature points to be taken as separate.

10. The procedure which is applied at claim 9, which is fully comprised of the
speed which is of type step computing and also error estimation is done for each
and every features and atlast by combining all the values to make the speed value as
single value and error of speed can be measured easily.

11. The procedure which is applied at claim 2 is where the important feature points
are to be taken as tops and bottoms of the plurality of the characters which is to be
detected on vehicle’s license plate.

12. The procedure which is applied at claim 1 is that they will be using one of the
camera positions as a constant variable type.

13. The procedure which is applied at claim 1, which is used to capture at least one,
sort of the video when vehicle cross the path of region

PATENT 3:

TITLE : VIDEO SPEED DETECTION SYSTEM

APPLICATION NUMBER: 11/970,371
PATENT NO : US 2008/0166023 A1
INVENTORS: Jigang Wang
DATE OF PATENT PUBLISHED : Jul. 10, 2008
COUNTRY : UNITED STATES
ABSTRACT: Here the system is used for measuring the speed by use of video
sensing. This system is comprised of tracking the camera which is arranged at time
shaped with higher precised video types which is also processing unit which is used
for analyzing sequences for real time for detecting the vehicles in motion and detect
its speed fro time user distance method. This system is known for detecting the
moving the vehicles where in each image frame is derived vehicle positions for
mapping the established from reference markers which are made on road. So the
bodies of the invention is going to be facilitate when position of the vehicle and
also the driftness which is calculated without any help of any manual or by
calculating of the position of the camera which is located in space when compared
with respect to the road area.
.
CLAIMS:
1. Here the system is used for monitoring the objects which is comprised of :

19
  a tracking camera which is created and kept it for providing an
plurality typed stamped images for measuring the area.
 a processing unit, coupled to the tracking camera, that is configured to
receive the time stamped images and to process the images to
determine from the time-stamped images a speed of a vehicle passing
through the measurement area.

2. The procedure which is applied at claim 1 where it has high resolution type of
camera coupled images which in response for forming the capturing of the image
which sends the signal for processing unit.

3. The model used as in claim 1 where confined model will be aiming for printing
the data, time and speed for the reticulation of the vehicle for getting higher
resolution image from the camera.

4. The model of claim 2, it is a processing unit which is confined to the receiving

the trigger speed value used for user of the system.

5. The claim 4 suggest that processing unit has the configured of comparing the
speed and triggering value of the vehicle for sending the signal with higher
resolution when the vehicle crosses the speed limit.

6. The system for claiming 4 which is used for system which is fully confined a
storage type and even it comprises for the recording of the time stamped images of
the vehicle where the speed exceeds the trigger point as vehicle crosses the path of
the measurement area.

7. The procedure which is applied at claim 1 which is used for time stamped images
and storage devices too which is to be constructed, and this processing unit is
further confined to receive the record time stamped images of the vehicles and used
for knowing the speed of recorded stamp of images.

8. The procedure which is applied at claim 1, which is also the processing unit
which is later confined for providing GUI to the user of the system.

9. The procedure which is applied at claim 1 is that GUI is configured for knowing
the assigned dimensions which is to find the measurement area which is known for
a user of the system.

10. The system which is used for claiming 8 is that GUI configuration is confirmed
which provides live link fir tracking camera for measuring the area which is used in
live video.

11. The procedure which is applied at claim 1, it is using the processing unit to find
out the speed update continuously for the speed of the vehicle and for displaying
the driftness of the vehicle at the position of vehicles which are seen in that video.

20
12. The procedure which is applied at claim 1 is that they used GUI which is used
to configure the displays of the recorded video timings which is to be stamped of
the vehicle.

13. The system of applied which is claim 1, which is used for the processing unit
which is configured for allowing all the users to know about everyone the all users
of the system for stepping through the recorded images and to find out the frames at
which the automobile crosses two intrusion lines which are known at some known
distance.

14. The procedure which is applied at claim 1, which is used to processing units
which configured for the construction of the configured images of the measurement
area which is used for the plurality of the time stamped.

15. The system for claim 15, which has the processor in it which is configured to
create the images to be rectified for the area which is used to determine the
projective mapping that from four different correspondences.

16. The system which is used for claiming for 16, which the processing unit which
is further configured for the map coordinates of at least one of the plurality of time
stamped images for knowing the actual images for measuring the area of projective
mapping.

17. This procedure which is applied at claim 1, wherein the processing unit which
can be known for further configured for the extraction for background images for
the measurement area for the plurality of the images stamped time wise.

18. For procedure which is applied at claim 1, which there they have taken
processing unit which is further configured for detecting and identifying the
vehicles for measuring the area.

19. The procedure which is applied at claim 1, which is also known as processing
unit which is configured to use for detecting the vehicles can be segmented of the
rectified image and which will be having large deviations in the rectified type of
images

21
PATENT 4:

TITLE : Video-based vehicle detection and tracking using spatio-temporal maps

APPLICATION NUMBER: 12/987,854

PATENT NO : US 8,358,808 B2
INVENTORS: Yegor Malinovskiy, Shoreline, WA
YINHAI WANG, SEATTLE WA(US)
Yao-Jan Wu, Charlottesville, (US)
APPLICANTS: NIL
DATE OF PATENT GRANTED: Jan. 22, 2013
COUNTRY : UNITED STATES PATENT
ABSTRACT: This system is known for detecting the moving the vehicles where in
each image frame is derived vehicle positions for mapping the established from
reference markers which are made on road. The information about the time is taken
from every time and days which we have related for every and every single frame.
This is a model which will allow many users of the system for setting some of the
speed which triggers and also it is used to keep every single video clip
automatically of the vehicles which will cross the triggering driftness, which will be
recorded of vehicle passing the zone of interest which is passed by the speeder
vehicle which was marked by us a reference. For the improved system and methods
which are used for finding the driftness of the automobile on the multilane highway
which is precisely used of sequence of the real time of the images. So from the
various types of the vehicle position detected and by which speed can also be
estimated which is fully based upon geometric data which is available of some
vehicles but unlike some characteristics like end points of the license plate
characters and when the points or characters of the license plate are totally intact.

CLAIMS:
1. Here the system is used for monitoring the objects which is comprised of :
 a tracking camera which is created and kept it for providing an
plurality typed stamped images for measuring the area.
 Dealing with a first correlation number which is helpful in describing
the features which are same for full size templates.
 . For saving the first correlation numbers and the location of that
features of that using frame.
 Wherever for searching the frames for current and upcoming for the
features matching for one scaled templates.
2. For procedure which is applied at claim 1 where it has high resolution type
of camera coupled images which in response for forming the capturing of the
image which sends the signal for processing unit.

22
3. The model used as in claim 1 where confined model will be aiming for
printing the data, time and speed for the reticulation of the vehicle for getting
higher resolution image from the camera

4. The procedure which is applied at claim 1 where the steps are fully
comprised of the basic height of the character which is to look up for the
characters on the number plate.

5. The procedure which is applied at claim 1, where it is fully comprised of the

steps which will help to perform the normalization of transforming by the
image prior for determining the steps.

6. The procedure which is applied at claim 1, which is fully having the distance
from the camera to which the character used for pairs of ray type which is
totally depended upon distance between the characters of the points.

7. The method of applying claim 6, later on which is comprised of the step

which is used for creating their own trajectory which is shown as the real
world coordinate points of the real imagining world.

8. The procedure which was used for claim 1, which is further comprised, if the
detected speed of average of the automobile might be cross a value of
decided earlier to find threshold driftness of the vehicle, which helps to
perform speed estimation which is also called to be on a realistic
decompressed type for the data used in all input videos.

9. The procedure which was used for claim 1, where the camera is attached in
at top of the fixed and constant makings.

10. In procedure which was used for claim 1, where the camera is fixed on the
vehicle top for capturing the video.

11. The type of processor which was used to execute in the processor with
some instructions for performing the same method which was used in claim
1, stating that instruction being kept preserved are computer read able only
type.

12. This is a computer programming which is comprised for the non transitory
computer type which is readable media type which will encode instructions
for claiming the method 1

23
 13. This is also the type of system of moving vehicle which includes of tracking
device, it includes memory storing and for performing for claiming 1 and
includes processor too for execution of the instructions

PATENT 5:

TITLE : Video-tracking for video-based speed enforcement

APPLICATION NUMBER: 13/611,718
PATENT NO : US 8,971,573 B2
INVENTORS: Wencheng Wu, Webster, NY (US)
DATE OF PATENT GRANTED: Mar. 3, 2015
COUNTRY : UNITED STATES PATENT
ABSTRACT: This is a method for tracking the moving vehicle which includes
detection of the vehicle by acquiring the series of temporal related images related
frames. From initial frames the detection includes the reference features which are a
total representative of the vehicle. So this method further extends setting the
reference lines as full size templates. In this method there is use of tracking of the
vehicle by searching a current frame for the features of the full size templates. This
tracking further includes setting as an updated templates one of the full size and
scaled templates which are the closet which are matching the features.
CLAIMS:
1. There are some of the methods which are used for detecting the speed of the
motor vehicles, these are the steps which are comprised of:
 providing a basic type which is comprised of one type of the
camera’s videos where they can be used for capturing a number of
series of successive where its separated by time which images will be
having a features of physical type.
 Which is used for knowing the location of features within which each
of the images are time separated;
 based on the known geometric parameter of the feature and a
geometry of the camera relative to the vehicle, estimating, for each of
the time-separated images, a real-world spatial coordinate position of
the feature; and
 based on the estimated real-world spatial coordinate positions of the
feature in the time-separated images and a capture time of each of the
time-separated images, estimating a speed of the moving vehicle.

2. For procedure in processing the claim 1 which includes the get information
of all vehicles, for knowing the characteristics which are between for
selecting the all video parts in along a way for the position of the vehicle’s

24
 characteristics frames; which is used for finding the driftness by finding out
the distance.

3. For procedure in of processing the claim 1 where it includes all

characteristics of selecting any one of position of a number plate detection.

4. This procedure for claiming 3 its use of scaling the full size template which
is used for pre determining of amount which is used to generate the
templates.
5. The procedure for claiming 4 is used to predetermine the amount which is
used to scaling for using the calibration data, and also helps to know the
exact position of that characteristics and finding the driftness for the present
traffic conditions which is always checked for tracking features.
6. The method of processing claim 1 for selecting which frame with full size
templates includes:
 Dealing with a first correlation number which is helpful in describing
the features which are same for full size templates.
 . For saving the first correlation numbers and the location of that
features of that using frame.
 Wherever for searching the frames for current and upcoming for the
features matching for one scaled templates
 For the computational second correlation which is used for describing
between features and scaled templates.

7. The proposed methodology for claim 6 where it is used for setting the
updated templates which includes:
 matching the first and second correlation numbers;
 This is used to determine for higher ones of first and second
correlations numbers.
 setting as the updated template the one of the full-size template and
scaled template corresponding to the higher correlation number

8. The proposed model for claim 1 is comprised of :

 cropping a region around the feature; and,
 setting the cropped region as the full size template

9. This is a computer programming which is comprised for the non transitory

computer type which is readable media type which will encode instructions
for claiming the method 1

25
 10. This is also the type of system of moving vehicle which includes of tracking
device, it includes memory storing and for performing for claiming 1 and
includes processor too for execution of the instructions.

PATENT 6:
TITLE : Integrated still image, motion video and speed measurement system
APPLICATION NUMBER: 12/236,288
PATENT NO : US 2009/0079960 A1
INVENTORS: Jiyoon Chung, Aurora, CO (US)
DATE OF PATENT GRANTED: Mar. 26, 2009
COUNTRY: united states patent
ABSTRACT: Here the system is used for measuring the speed by use of video
sensing. This system is comprised of tracking the camera which is arranged at time
shaped with higher precised video types which is also processing unit which is used
for analyzing sequences for real time for detecting the vehicles in motion and detect
its speed fro time user distance method. This system is known for detecting the
moving the vehicles where in each image frame is derived vehicle positions for
mapping the established from reference markers which are made on road. So from
the various types of the vehicle position detected and by which speed can also be
estimated which is fully based upon geometric data which is available of some
vehicles but unlike some characteristics like end points of the license plate
characters and when the points or characters of the license plate are totally intact in
this system.

CLAIMS:
1. A computer-implemented type for detecting the speed of the vehicle from
motion vector information which includes:
This method for tracking a vehicle moving is done by some methods which
include:
 Detecting a vehicle includes:
 acquiring a series of temporal based frames of the images;
 for a basic frame, location of a feature for a reference representative of
the vehicles;
 setting the characteristics as a reference for using as full size template ;

26
2. For claiming procedure which was used at claim 1, which we have used for
input type datasets type of motion vectors used in the method of compressing
the input data set.

3. The procedure which was used at claim 2, for estimating each and every
frames driftness and finding the average speed of the vehicle which is done
by combining the vector lengths which are measured.

4. The procedure of 3, which we have used the mixed type of vector inputs of
different lengths which is of an average type and is an order of statistic type.

5. The procedure which is meant for claim 1, by using where the motion vector
form of the cluster is regular form to delete the input vectors which are
considered for never to correspond for expected vehicle trajectory.

6. The procedure which was used for claim 1, where received dataset in form of
videos is fully compressed form of data video set and further we can identify
the motion vectors which contains motion input data which we have taken it
from the video data set which is now compressed.

7. For the method of processing the claim 1 where it includes the feature of at
minimum of one position at number plate.

8. The procedure of claiming 3 its use of scaling the full size template which is
used for pre determining of amount which is used to generate the templates.

9. The procedure for claiming 4 is used to predetermine the amount which is

used to scaling for using the calibration data, and also helps to know the
features as position and driftness of the average vehicle of the present traffic
condition which is always checked for tracking features.

10. The procedure which is applied at claim 9, which is fully comprised of the
speed which is of type step computing and also error estimation is done for
each and every features and atlast by combining all the values to make the
speed value as single value and error of speed can be measured easily.

11. The procedure which is applied at claim 2 is where the important feature
points are to be taken as tops and bottoms of the plurality of the characters
which is to be detected on vehicle’s license plate.

12. The procedure which is applied at claim 1 is that they will be using one of
the camera positions as a constant variable type.

27
 13. The procedure which is applied at claim 1, which is used to capture at least
one, sort of the video when vehicle cross the path of region.

14. The procedure which was used for claim 1, where we have got the video
input data as dataset which is fully comprised of fully uncompressed data
with a live stream, and wherein received data and motion vectors are to
identified which are usually created during the compression of the video
frames.

CHAPTER 4: Problem definition, Initial Design using Design

Thinking approach, Flowcharts AND Algorithms

4.1 Problem Definition:

Vehicle tracking and speed measurement is a process of locating a moving vehicle and
obtaining speed of vehicle using camera. Record vehicle in video sequence frame by
frame from traffic camera is most wanted application to improve traffic and vehicle
tracking performance. Using This technology we can achieve the number of applications
.those are traffic control, traffic monitoring, traffic flow, security etc .we recommend
Video and image processing has been used for above applications, traffic control ,traffic
flow monitoring traffic conditions ,it also analysis the traffic surveillance .Our aim is to
develop automation system that can accurately identifies and track the speed of any
vehicles that showed in video footage taking video information as input we can achieve
this less expensive and less maintenance.
Objectives : To suggest a video based speed measurement system that can cover a road
which vehicle can pass through it or a multi-lane road with less appliance and
maintenance costs in comparison with sensors.

4.2 Initial design using Design thinking approach:

Dataset creation : Video of moving vehicle on highway/streets road recorded from a
camera particular speed for experimentation and evaluation .Taking these videos we can
implement project.

Software requirements:
Python based Computer Vision, image processing and Deep Learning libraries are used
for development and experiment of project .Tools used are python ,libraries used are Open
CV.
A high-level commercial purpose programming language. This language constructs an
object-intended approach which is going help software developers to write a logical code
for different scale objects .Python is changing typed and garbage stored.It supports
multiple programming patterns including structured .

28
Applications of pythons:

 Making apps using python code .

 Different types of software development.
 Some mathematics applications .
 we can write system documenting.

 We may use python to link database management system. In this, we can read and
modify files .
 Python can be used to hold large data and perform compounded mathematics and
its application.
 It works on different platforms such as Windows, Linux. As we know.
 Python has a simple syntax as same as English language .Both are easy to learn.
 It has syntax that allows software programmers to write programs with less line
than some other programming languages.

 Python runs supporter system .that is, program can be executed as soon as it is
written. prototyping could be very fast.
 It can be treated in a systematic way and an object intended way and a practical
way.
 Using python we can write integrated development environment useful when
managing larger collections of files.
 It was designed for understandability and has some similarities to the English
language with influence from mathematics.
 It uses new-lines to complete a command as opposed to other programming
languages like c/c++,which often uses semicolons or parenthese.
 It depends on spaces on the beginning of the code.it uses whitespace .Other
programming languages use curly brackets for that purpose.
 It can used in a server to generate web applications.

4.3 Tools and Technologies used :

OpenCV:

OpenCV supports different kind of programming languages namely c++, java,

python and many.

OpenCV-python is a library python bindings designed to solve the computer

vision problem.

Haar cascade: Object identification using Haar feature based cascade classifiers is
most effective method in today’s world . Object Detection using a Boosted Cascade
of Simple Features. Two types of images are there to find object .one is positive
image and other one is negative. it is used to identify the objects in the video.

4.4 Flow chart:

29
30
Object detection : OpenCV is the open source library for computer vision, which is field
of AI that trains computers to clarifies and understand the visual world. Using digital
pictures and videos machines can accurately identify and clarify the object. machine
learning, this allows software application to become accurate at predicting outputs.
Image processing , in image processing we can achieve low quality images to the high
quality image.it also important in real time operations.
Using opencv we can process images and videos to recognize the objects ,as well as
faces of humans. Object identification is a computer technology respond to computer
vision ,image processing, and deep learning that deals with detecting examples of
objects in a images and videos .This was done by haar cascade.
Haar Cascade is a machine learning based approach where positive and negative images
are used to train the classifier. We use two type of images .one is Positive images, these
images contain the images which our classifier to detect. Second is Negative Images,
these Images of everything ,which do not contain the object to detect in picture.
Vehicle tracing: Our aim is to record the coordinates of the moving object and figure out
that object in the video. We would want our project to identify the moving object in a
video . The moving car is detected and a bounding box is created surrounding the car.
31
There were several techniques to solve that problem. we can train a deep learning model
for object identification and we took a pre trained model and fine-tune it on our data.
Insertion of reference lines:
At first we create a data sets our own. in that we mark reference line for our convenience
finding a speed of moving vehicle in programming code. these are intrusion lines. Three
reference lines are used . These reference lines are about same distance such as 20m
.passing through these reference lines speed of the vehicle can be observed.

Speed Estimation: Speed can be estimated when a vehicle enters the reference line and
region of interest .Objects position is acquired at distinct locations due to the time related
sampling rate. The detection distance is directly proportional to the speed of the vehicle
and inversely proportional to the camera’s frame rate. For successful detection the vehicle
should be within the several intrusion lines or region of interest. Distance vector of the
intrusion lines are taken into consideration set of possible positions can be computed.
These positions are indicated detection distances within a vehicle traveling at a different
speed should appear in .Set of different possible positions is computed using detected
movement pattern vector.

We take datasets as reference .we made video of moving vehicle at different speed .After
getting these video in program we can extract in the form of MP4.we may then measure
the distance using pixel values. Detection of vehicle passing through the reference lines.
Measure the time taken by the vehicle to pass through adjacent reference lines. We know
the relation between the time, speed and distance .speed is equal to distance divided by
time. We have measured driftness of the vehicle using time.
Algorithm:
 First we create datasets .these datasets are created on taking videos of moving
vehicle at different speed.
 Before taking a datasets, we must mark/insert 3 reference lines at same distance
like 20m. Because we show a speed of vehicle at these intrusion line.
 Datasets are created in different ways, different angles and different speeds such as
40kmph, 70kmph and 100kmph.
 These datasets are taken as input to the program as MP4.
 Then we can write a code for object detection means vehicle detection in python
programming language. These are done using opencv library.
 While detecting object, we add picture as jpg for detection.
 If we detect a vehicle in a video frame then we can write a code for tracing a
vehicle in a video.
 Measurement of distance using pixel values.
 After getting a region of interest we can detect a vehicle passing through reference
line.
 Measurement of time taken by a vehicle to pass through adjacent reference lines.
 Measurement of driftness of the automobile using time and measurement between
the lines which are taken as reference.

32
  It will show all the results approximately as real.

CHAPTER 5: Result and Discussion:

 In this project we have discussed the issues of vehicle to find out the speed
depending upon videos which is taken by camera which is fixed in urban
places. Here we have introduced a model which is dependent on selecting
and detecting and tracking of the instinctive characteristics which is situated
in every single vehicle license plate region. So in this system we have almost
tested with every single type of the vehicles that is bike, cars and heavy
vehicles too with full high quality, with unlimited data sets created by us and
detected on multiple road lanes. , which we have stated in our paper is all
real values which we have taken with higher accuracy system which is based
on many type of inductors loop detectors and these all are truth which is
labeled to license plate regions. So in our project we have detected license
plate region which yielded us the better results rather than taking the vehic

45.5 km/h (real 48.0 km/h) 43.5 km/h (real 44.6 km/h) 54.7km/h(real54.6 km/h)

43 km/h (real 43.0 km/h) 48 km/h (real 46.3 km/h) 39.9 km/h (real 41.2 km/h

33
 FIG(4)

 As we continued our work we are intending to identify the distance of

license plate from the points so that we can make higher yields for ex: for
dealing with the flow of the traffic, to identify stolen automobiles, etc. so one
more aspect of this work which helps in future is the implementation for
helping in platforms which will give permissions to all processors on local,
which also gets some of type parallel processors which used in GPUs, for
which it requires common bandwidth used for communication.
 The project is depended upon speed detection by seeing videos which gives
us the probability density function of automobile which crosses with some
speed based on its moving way or path input. So in this work we have
proposed a new type which we have used by arranging some of the intrusion
lines and taking it as data input frames which we have captured
simultaneously by using the camera which we have fixed on upper hand of
the road with least minimal installation and by using Smartphone high pixel
camera with each frame rate of 50fps and 40fps respectively. So when we
have taken our project details and results, the error rate of the vehicle speed
is at 1% and 2% respectively, for vehicle which is moving with speed of 60-
80km/hr. So we can reduce error rate when we increased number of intrusion
lines. So further the output values which we got are in range when compared
with pdf, which says or yields our project confidence in this project. So we
also got to know the values of the pdf also will reduce when we increase the
camera frame rate. So in this project we have checked around 80 different
speed and inputs and the speed we got we have compared to the speed where
vehicle passing the zone of intrusion lines. So atlast we have found that the
speed which we detected and the actual speed of vehicle are almost same and
related. So we have made this relation for the project which we have used for
majority of the heavy vehicle like cars, buses we have equipped with some
speed regulators so we can follow the speed limits issued. So we can
conclude that the video based speed measurement of the vehicle will yield us
with high accuracy and robustness. So we in future work should be focusing
on using the map so deeply so that we can arrange the intrusion lines
wherever the need of it.
 As a proof we have done experiment on the national highway using a GPS
installed in it and camera is installed in it. So we have made recording by
using the pictures which has around 50fps input rate of each frames. So
when automobile is crosses the area for finding out passed at valuable speed,
maintaining the ground truth. So our main objective of this project is to

34
 check our project used measuring values for the driftness of the vehicle and
find it most precise by using the most of the intrusion zone formed.
 Region which must be detected which we have created by using many
intrusion lines for every single and every input we made as line. The vehicle
passes the intrusion lines using the model analytically. This project when
computed by using the pdf of the vehicle speed which is used by movement
of the vector input, intrusion lines and the sampling rate and time of the
camera. So the proposed model which we used is so compatible to older
methods and we can improve the deflection and precision because all the
ideas we measure as displacement for some period of the time based on the
frames too. By using this proposed model we can find the speed of the
vehicle by using some video frames.

CHAPTER 6 : Conclusion and scope for future work

CONCLUSION:
The designed model which we used for speed detection system has a capacity of
continuously keeping an attention of the speed of vehicle coming towards the
region of our interest. The sensors which are acquired are kept for the vehicles to
the nearest range around some meters. So the output which we get is higher precise
when other vehicles are not in the movement. So the vehicle when passes the
intrusion lines will be displayed on the LCD. With each automobile passing the
zone with more speed sensor will detect and camera will save the memory. Then
that image will be sent to all network servers that’s internet where by seeing the
image the number plate will be detected. So this plate recognition is not that
accurate and needs some changes. The number plate detection is bit difficult due to
different font style and size to detect it. It will be easy to identify the number plate
if it’s of similar style which us shown in result.
So we can conclude that:
 The proposed project involves the speed detection of the vehicle which we
have been used by using the method of processing of the images. So
automobile driftness detection is done by utilising some of the many types of
approaches which is done by movement of multiple vehicle estimation,
tracking of that automobile … these all results in higher accuracy.
 So this experiment we have carried on video samples taken on highway
which was utilized for this reason. So the multiple vehicle detection is
estimated by gradient edge detection and normal correlation cross type of
algorithm. Atlast driftness of the vehicle is determined in form of the

35
 tracking of these specific automobile features that too on depending upon
both length and rate of inner frames also.
 For performing our model project system for calculating by many measures
likely preciseness, error correction, higher precision, specific nature and
sensitivity. The vehicle tracking and speed detection gives with higher
precision.

Hence it is seen that by our proposed model we found that it has higher accuracy
than older running models. So atlast we can conclude that determination of the
driftness is measured by seeing all video type pictures is highly efficient and mostly
economical than nowadays used techniques like radar technology.

FUTURE SCOPE:

Our project aims that to support the vector type algorithm which can be utilized for
the speed estimation of the vehicle so that we can get higher estimation output. If
we take lower quality for all the video based inputs or vehicle number plate as input
which was unknown found correctly that’s why our future work or challenge is to
college the higher resolution type of dataset inputs which will help for finding the
license number of the vehicle which is for implementing driftness of the vehicle by
the algorithm which we utilized are very important measure to be considered into
consideration for developing the speed estimation of the vehicle.

So in future we find our projects having very high scope, demand and will be so
useful to implement in upcoming future world:

 Road safety:

So timely correcting and checking the vehicles which are travelling

with over speed will help to find out the vehicle and it helps to reduce the
very large percentage of the road accidents by taking a crucial action against
these type of vehicles.

 Automation in law enforcement:

The model because system is being automatically , it reduces the need

of the traffic police officers number which is needed to be deployed in
the real running world for the use of checking speed of the running of
the vehicles.

With very few types of the system which has new features will be easily used such
as vehicle will be detected by comparison by the stolen vehicles of the registered
automobiles and there is also system in the parking lounge or complex purpose
management systems which is used to recognize by guest vehicle.

36
Reference:
1) Lan J, Li J, Hu G, Ran B, Wang L. Vehicle speed measurement
based on gray constraint optical flow algorithm. Optik - Int J
Light Electron Opt 2014;125(1):289–95.
2) Gomes SL, de S Rebouças E, Neto EC, Papa JP, de
Albuquerque VHC, Filho PPR, et al. Embedded real-time speed
limit sign recognition using image processing and machine
learning techniques. Neural Comput Appl 2017;28(1):573–84.
3) Neto EC, Gomes SL, Filho PPR, de Albuquerque VHC.
Brazilian vehicle identification using a new embedded plate
recognition system. Measurement 2015;70:36–46.
4) Buch N, Velastin SA, Orwell J. A review of computer vision
techniques for the analysis of urban traffic. IEEE Trans Intell
Transp Syst 2011;12(3):920–39.
5) Kaya S, Kilic N, Kocak T, Gungor C. A battery-friendly data
acquisition model for vehicular speed estimation. Comput
Electr Eng 2016;50:79–90.
6) Luvizon DC, Nassu BT, Minetto R. A video-based system for
vehicle speed measurement in urban roadways. IEEE Trans
Intell Transp Syst 2017;18(6):1393–404.
7) Banerjee S, Dhar M, Sen S. A novel technique to detect the
number of ground vehicles along with respective speed of
each vehicle from a given video. In: 2018 emerging trends
in electronic devices and computational techniques (EDCT);
2018. p. 1–6.
8) Zhiwei H, Yuanyuan L, Xueyi Y. Models of vehicle speeds
measurement with a single camera. In: Proceedings 2007
international conference on computational intelligence and
security workshops; 2007. p. 283–6.
doi:10.1109/CISW.2007.4425492.
9) Madasu VK, Hanmandlu M. Estimation of vehicle speed by
motion tracking on image sequences. In: Proceedings 2010
IEEE intelligent vehicles sym- posium; 2010. p. 185–90.
doi:10.1109/IVS.2010.5548051.
10) He XC, Yung NHC. A novel algorithm for estimating
vehicle speed from two consecutive images. In: Proceedings
2007 IEEE workshop on applications of computer vision;
2007. 12–12
11) Maduro C, Batista K, Peixoto P, Batista J. Estimation of
vehicle velocity and traffic intensity using rectified images. In:
Proceedings 2008 15th IEEE international conference on

37
 image processing; 2008. p. 777–80.
doi:10.1109/ICIP.2008.4711870.
12) Jeyabharathi D, Dejey D. Vehicle tracking and speed
measurement system (vtsm) based on novel feature
descriptor: Diagonal hexadecimal pattern (dhp). J Visual
Commun Image Represent 2016;40:816–30.
doi:10.1016/j.jvcir.2016.08.011.
13) Cheung S-CS, Kamath C. Robust background subtraction
with foreground validation for urban traffic video. EURASIP J
Appl Signal Process 2005;2005(14):2330–40.
14) Brutzer S, Hferlin B, Heidemann G. Evaluation of
background subtraction techniques for video surveillance. In:
Proceedings 2011 IEEE conference on computer vision and
pattern recognition; 2011. p. 1937–44.
15) Hua S, Kapoor M, Anastasiu DC. Vehicle tracking and
speed estimation from traffic videos. In: 2018 IEEE/CVF
conference on computer vision and pattern recognition
workshops (CVPRW); 2018. p. 153–1537.
doi:10.1109/CVPRW.2018.00028.
16) Doan S, Temiz MS, Klr S. Real time speed estimation of
moving vehicles from side view images from an uncalibrated
video camera. Sensors 2010;10(5):4805–24.
doi:10.3390/s100504805.

APPENDIX 1:
SOURCE CODE:
import cv2
from tracker import *

# Create tracker object

tracker = EuclideanDistTracker()

carCascade = cv2.CascadeClassifier('myhaar.xml')
#cap = cv2.VideoCapture("VID_20210311_101556.mp4")
cap = cv2.VideoCapture("VID_20210311_101512.mp4")

fps= cap.get(cv2.CAP_PROP_FPS)

# Object detection from Stable camera

38
object_detector = cv2.createBackgroundSubtractorMOG2(history=100,
varThreshold=40)

d=0
i=0
speed=[None]*1000
id_1=[None]*1000
time_location1={}
time_location2={}
Center={}
ppm=83
pbl=136

while(cap.isOpened()):
ret, frame = cap.read()
if(ret==False):
break
height, width, _ = frame.shape
i=i+1
# Extract Region of interest
#roi = frame[275: 575,300: 800]
frame1=frame
frame=cv2.resize(frame, (240, 135))
#frame=cv2.resize(frame, (960, 540))
roi=frame[:]
# 1. Object Detection
mask = object_detector.apply(roi)
_, mask = cv2.threshold(mask, 254, 255, cv2.THRESH_BINARY)

gray = cv2.cvtColor(roi, cv2.COLOR_BGR2GRAY)

cars = carCascade.detectMultiScale(gray, 1.007, 40)
#cars = carCascade.detectMultiScale(gray, 1.2, 13)
detections = []
for (x, y, w, h) in cars:
# Calculate area and remove small elements
detections.append([x, y, w, h])
#cv2.rectangle(roi, (x, y), (x + w, y + h), (0, 255, 0), 3)

# 2. Object Tracking
boxes_ids = tracker.update(detections)
for box_id in boxes_ids:
x, y, w, h, id = box_id
d=d+1
Center[id]= x+w/2
print(Center[id])
if(Center[id]>71 and Center[id]<80 ):

39
 time_location1[id]=i/fps
print(".................11111111111111111...................")
id_1[id]=1

if(Center[id]>105 and Center[id]<115 and id_1[id]==1):

time_location2[id]=i/fps
t=time_location2[id]-time_location1[id]
speed[id]=pbl/ppm*3.6/t
print(".................22222222222222222...................")
if(speed[id]!=None):
cv2.putText(frame1, str(int(speed[id]))+"km/h", (x*8, y*8 - 15),
cv2.FONT_HERSHEY_PLAIN, 3, (255, 255, 255), 2)
cv2.rectangle(frame1, (x*8, y*8), ((x + w)*8, (y + h)*8), (0, 255, 0), 3)

frame=cv2.resize(frame1, (960, 540))

w, h,_=frame.shape
cv2.imshow("roi", roi)
cv2.line(frame,(int(960/2),0),(int(960/2),h),(255,0,0),2)
cv2.line(frame,(int(960/2-136),0),(int(960/2-136),h),(255,0,0),2)
cv2.line(frame,(int(50),0),(int(50),h),(255,0,0),2)
cv2.line(frame,(int(960-80),0),(int(960-80),h),(255,0,0),2)
cv2.imshow("Frame", frame)

cv2.imshow("Mask", mask)

key = cv2.waitKey(30)
if key == 27:
break

cap.release()
cv2.destroyAllWindows()

40