See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327337312

A Review on Vehicle Speed Detection using Image Processing

Article · November 2017

CITATIONS READS
10 9,684

1 author:

Balasubramani Ramasamy
NMAM Institute of Technology
42 PUBLICATIONS 169 CITATIONS

SEE PROFILE

All content following this page was uploaded by Balasubramani Ramasamy on 06 September 2018.

The user has requested enhancement of the downloaded file.

 A REVIEW ON VEHICLE SPEED DETECTION USING IMAGE
PROCESSING
Mangala A.G1, Dr. Balasubramani R2
1
Department of Master of Computer Application, N.M.A.M. Institute of Technology, Nitte
2
Department of Information Science and Engineering, N.M.A.M. Institute of Technology, Nitte

ABSTRACT II.LITERATURE REVIEW

Detecting the speed of vehicles is an important In image processing morphological operations
aspect for observing speed limitation law and highly experimented [16] in improving the
traffic condition. This paper gives review on appearance. To reduce the noise the MM is also
speed detection of vehicle using image applied it uses structuring element to probe the
processing. Given a sequence of real-time image and thereby useful information from the
video of traffic images, different approaches image can be obtained and noise can be reduced
like edge extraction, object tracking, motion while preserving the features. This paper is on an
vector technique, absolute difference, experiment in which four morphological
centroid method, background image operations are working to reduce the noise from
subtraction are useful in detection of vehicles the gray scale image and thereby enhancing the
speed. quality of the images. In the literal [7], authors
Keywords: Graythresh, erosion, dilation, introduce the first step towards developing the
grayscale. Speed Detection Radar, where he explains a new
approach in object detection technique, which is
I .INTRODUCTION “adaptive background subtraction”.
Image processing has been widely applied to Rad A. G. et al. [8] developed a system in which
traffic analysis for a variety of purposes. As they used video and image processing toolbox
traffic research field is very wide and it has many which calculates the speed of vehicle. It resulted
goals that include detection of queue, detection in average error of speed +7km/h and -7km/h.
of incident, classification of vehicles, and This system could operate on images with
counting vehicles. One of the most important of various resolutions and different video
these purposes is to estimate the speed of a sequences.
vehicle, a vehicle. Traffic congestion poses lot of Shedbalkar K. et al. [9] developed a speed
problems for people. Because of this, many estimation technique which was based on
accidents occur. To reduce this problem, new extended kalman filter for permanent magnet
approach has been developed for estimating the synchronous. System is developed in MATLAB
speed of vehicle. A radar technology was used to in SIMULINK model Blockset.
determine the speed on highways. But it has a Leite A.V. et al. [10] determined a way for
disadvantage of high cost. Then a lidar detector estimation of speed in induction motor with
was designed to detect the infrared emissions of sensor less control. Extended kalman filter was
law enforcement agencies lidar speed detection used as speed detection technique. This
devices and warn motorists that their speed is algorithm used reduce order state space model.
being measured. Its disadvantage is it has to be Kassen N. et al. [11] proposed a vehicle speed
held or placed at a static point. These drawbacks estimation technique which was reliable and
of speed detection techniques motivated to strong. This helps the user with driving guide and
develop new technique for that purpose. lets him not to join the traffic jam. This approach
is based on RF. This system gives accuracy of

ISSN (PRINT): 2393-8374, (ONLINE): 2394-0697, VOLUME-4, ISSUE-11, 2017

23
 INTERNATIONAL JOURNAL OF CURRENT ENGINEERING AND SCIENTIFIC RESEARCH (IJCESR)

100% for speed estimation and with accuracy of hulls that are found in a series of images. The
90% in typical streets. function bwdist computes the distance between
the centers of the convex hull.
III.TECHNIQUES USED
A. Proposed System [1]. B. Proposed System [2]
In this approach the sequence of images are first This presents a new Speed Detection Camera
converted into monochrome images. Then its System (SDCS) which is an alternative to Radar.
edges are extracted then detect the vehicle which It makes use of various image processing
may be tracked in various frames. Then the speed techniques on video stream in online that are
it calculated by comparing with ground truth. captured from single camera or offline mode,
Sequence of steps used are: which makes it capable for calculating speed of
1. Reading the video: Record the video in the vehicle. Radar limitations are overcome here by
Matlab workspace, which then must be read offering an in-expensive approach with the same
using (aviread) function which read the video accuracy or better. Phases of SDCS are divided
from Matlab workspace. into 4 phases;
2. Convertion of the Video Frame to an Image: 1. Object Detection: Here an hybrid algorithm is
The function frame2im in MATLAB converts the used which will detect the moving object. It is
single movie frame into an indexed image and an based on the combination of adaptive
associated color map. background subtraction technique and a three-
3. Computing the Images Absolute difference: frame differencing algorithm.
The function imabsdiff subtracts each element in It has 3 steps
current image from the corresponding element in i. Constructing the motion matrix
the previous image and returns the absolute ii. The masked subtraction
difference in terms of array elements. iii. Generating new background threshold matrix
2. Objects Tracking: Frame by frame tracking of
4. Converting the image to Black and white: The object which has been detected is very difficult
image is converted to black and white by using task but at the same time it is important. It is the
the function im2bw. It converts an image to significant part of the systems as without object
binary image, based on threshold, the function tracking, the system will not be able to extract
graythresh is used to compute threshold cohesive temporal information about objects and
automatically which computes a global any higher level behavior analysis would not be
threshold. possible. On the other hand,
5. Labelling the objects in the Image and Finding tracking will be difficult if the foreground object
the Center of the vehicles to keep segmentation is inaccurate because of shadows,
track of them: All the components in the image reflectance and occlusions.
must be marked, which helps to determine the Object tracking has 3 phases;
number of labeled objects that is found in i. Object segmentation
numobjectsbase parameter, this step is related ii. Object labeling
with another step which will measure all the iii.Object center extraction
properties of the image components. 3. Speed Calculation: After tracking each object
6. Convex Hull Extraction: Then the moving in video the next step is to save the frame number
edges are filled which would appear as solid that the object entered the scene at (Fr0), and the
moving blobs. We use a convex hull which frame number that the object left the scene at
characterize the blobs, which would approximate (FrN), then speed calculation can be carried out
the contour of the vehicles. by calculating the number of frames consumed
7. Bounding Box Extraction: To obtain scaling by the object to pass-by the scene and as the
information directly from the image known duration of each frame is known, the total time
geometric relationships in the images is taken by the object to pass-by the whole scene
exploited, which is done by constructing a can be calculated.
bounding box to enclose the convex hull. 4. Capture Object's Picture: The best position to
8. Geometric analysis: By keeping the direction capture a picture with good resolution is when
of motion of each vehicle fixed, the best fit line the object is at the center of the screen. So at that
is computed through the centroids of the convex time system stores the captured picture as current

ISSN (PRINT): 2393-8374, (ONLINE): 2394-0697, VOLUME-4, ISSUE-11, 2017

24
 INTERNATIONAL JOURNAL OF CURRENT ENGINEERING AND SCIENTIFIC RESEARCH (IJCESR)

frame. Then the system marks the targeted object 2. Vehicle velocity estimation: Once the image is
in the frame so that there will be some difference extracted and segmented into blocks, the motion
between it and other moving objects within the vector technique is applied to calculate the pixels
scene. changes among the two blocks to measure the
C. Proposed System[3] speed of the moving vehicle. The motion vector
In this approach 2 extracted images are selected is applied with the vector valued function to
to apply the motion estimation process in the demonstrate the vehicle speed detection
developed MATLAB algorithm. Standalone algorithm for the video from surveillance
images is segmented into 16 × 16 small blocks cameras.
using the division technique. Then the blocks are 3. Vector-valued function for vehicle motion
extracted from video coding to be compared with velocity: This is to verify the respective number
its respective image in current image and the of changes in blocks within two consecutive
previous image. This blocks are compared to images. The input of a vector valued function
detect the changes in pixels which is used to may be a scalar or a vector where as the output is
estimate the velocity of the respective moving a vector. When vector valued function is applied,
vehicle. the vehicle speed can be determined.
The flowchart is shown in Fig1 4. Digital Video Recorder (DVR) card setting
estimation: If we want to associate the sequence
images with vector valued function algorithms, a
digital video recorder (DVR) card is required. It
is installed to capture the video from the camera
and save into a particular folder in the hard drive.
D. Proposed System[4]
It presents a method which determines the
vehicle speed in accident prone areas by using
the video frames that are captured from the
camera which are fixed on the road. It works as
follows:
1. Frame Differencing Operation: One frame is
taken as the reference frame and then two
different frames are taken to calculate the
absolute difference between those two frames
with the reference frame as in Fig2.

(a)Frame X

Fig1. Vehicle speed detection algorithm

organization

The successive steps are

1. Video image processing: Block extraction and (b)Frame Y
subtraction technique is applied into region of Fig 2: Result of Frame differencing being
interest instead the complete video sequence or performed on (a) Frame X and (b) Frame Y
images. with background frame

ISSN (PRINT): 2393-8374, (ONLINE): 2394-0697, VOLUME-4, ISSUE-11, 2017

25
 INTERNATIONAL JOURNAL OF CURRENT ENGINEERING AND SCIENTIFIC RESEARCH (IJCESR)

2. Thresholding Operations: Image E. Proposed System[5]

Thresholding is a method in which image is The core technique used to determine vehicle
effectively partitioned into a foreground and speed in this approach is centroid method. The
background. This is one of the segmentation centroid value of vehicle is calculated when it is
types which isolates objects by converting them at position A. After traversing distance, when it
from grayscale images into binary images. It is reaches at position B find the centroid value of
converted into binary image because it easy to vehicle again. From these values, distance is
perform the motion operations on the binary calculated and from this the speed is derived in
image than the grayscale image. meter per second.

3. Morphological Operations: This is performed Steps are:

to remove all the isolated points that are observed 1. Image from camera: Input image is taken as
in the thresholding operations. After removing the images captured from CCTV camera,
the isolated points image is left with some installed at junction or freeway.
disconnected components. All these 2. Reading Template Image: Individual vehicle
disconnected components are connected together images are stored in database. From that, read
than calculating centroid of all these template images of vehicle.
disconnected component as in Fig6. Then 3. Convert into Gray Scale: Both the input image
morphological closing is performed to connect and template image has been converted into gray
the vehicle disconnected components together scale images as it gives better result on
for Frame X and Frame Y as in Fig3. comparison.
4. Dividing into blocks: Input image is divided
into blocks. Same sized blocks are extracted
from each column and from each row with each
image having size of block.
5. Matching of blocks: Template image is
matched with each individual block by using co-
relation function with matching factor as 9. If
both are matched then the information about
matched block is retrieved.
6. Computing coordinates value of matched
block: After matching of block, coordinates
value of matched block has been calculated.
Block has a shape of rectangle. So it will have
Fig 3: Morphological operations are performed four coordinates. First two coordinates have been
on (a)Frame X and (b) Frame Y to clean up all calculated by finding the starting and ending
the isolated points. coordinate of block in first row. Other two
coordinates are calculated by adding the size of
block to that coordinates.
7. Computing centroid value: Area of block is
calculated followed by calculating the Centroid
value of matched block.
8. Read vehicle image after traveling distance:
After traveling distance, vehicle image has been
taken from camera.
9. Converting into Gray scale: Then the image is
converted into gray scale for comparison.
Afterwards it is divided into blocks. Again
blocks are matched with template image and
information about most matched block is
retrieved.
Fig4 : Morphological closing is performed on
(a) Frame X and (b) Frame Y to join the
disconnected vehicle components together.

ISSN (PRINT): 2393-8374, (ONLINE): 2394-0697, VOLUME-4, ISSUE-11, 2017

26
 INTERNATIONAL JOURNAL OF CURRENT ENGINEERING AND SCIENTIFIC RESEARCH (IJCESR)

10. Computing coordinates value of matched iii. Tracking: Used to analyzing the position,
block: After traveling distance, coordinates value velocity and direction of moving object.
for matched blocks is calculated as in step 6.
11. Computing centroid value: Centroid value is IV.CONCLUSION
calculated in the same way as in step 7. 1. This paper gives review on vehicle speed
12. Calculating distance: Distance traveled by detection technique using different approaches.
vehicle either in horizontal or vertical direction, 2. Different approaches are edge extraction,
calculated by using distance formula. This object tracking, motion vector technique,
formula has been applied on centroid coordinates absolute, centroid method and background image
for two most matched blocks with template subtraction.
image. 3. The processing is done in MATLAB.
13. Notify time from CCTV camera: Time taken 4. By using any of these methods, traffic can be
by vehicle to travel distance is notified from controlled and vehicle speed detection will be
CCTV camera. maintained.
14. Calculating speed of vehicle: Speed of
vehicle is then calculated by distance traveled by REFERENCES
vehicle in particular instant of time. It is [1] Ammar Awni Abbass University of Baghdad
calculated by using mathematical concept of "Estimating vehicle speed using image
speed, time and distance. processing", AL-Mansour Journal / No.14/
Special Issue /( Part Two) 2010
F. Proposed System [6] [2] Osman Ibrahim, Hazem ElGendy, and
It proposed a real time to detect moving vehicles Ahmed M. ElShafee, Member, IEEE " Speed
which violate the speed limit. It makes use of Detection Camera System using Image
digital signal processing for estimating the speed Processing
of the moving vehicles. Techniques on Video Streams ", International
Journal of Computer and Electrical
1. Partitioning the road into disjoin region: In Engineering, Vol. 3, No. 6, December 2011
order to reduce the computational cost, time and [3] S. S. S. Ranjit, S. A. Anas, S. K.
the problems occurred in measurement that are Subramaniam, K. C. Lim, A. F. I. Fayeez, A. R.
caused by multiple vehicles moving side by side Amirah," Real-Time Vehicle Speed Detection
or one after each other, the road area is Algorithm using
partitioned into disjoint regions. The parts of the Motion Vector Technique ", Proc. of Int. Conf.
image that do not belong to any region are not on Advances in Electrical & Electronics 2012
considered in the calculations. Each region is [4] Siddharth Jhumat," Vehicle Speed
processed separately to reduce the computational Estimation in Accident Prone
cost. Areas using Image Processing ", International
2. Binary image generation: The speed Journal of Advanced Research in Computer and
measurement is performed in binary image in Communication Engineering Vol. 3, Issue 5,
which each pixel is transformed into either “1” May 2014
or “0” according to its motion information. To [5] Pratishtha Gupta, G N Purohit, Manisha
binarize the incoming input image and only Rathore," Estimating Speed of Vehicle using
detect the moving pixels, two different Centroid Method in MATLAB ", International
techniques are used: Journal of Computer Applications (0975 – 8887)
i. Inter-frame difference and Volume 102– No.14, September 2014
ii. Background subtraction. [6] N.Arun Prasath1 G.Sivakumar2
3. Tracking the moving object: The sequential Mr.N.Kumaresan3," Vehicle Speed
approach to track the moving objects and to find Measurement and Number Plate Detection using
the velocity of objects are as follow. Real Time Embedded System ", Network and
i. Segmentation: Process which separates Complex Systems
multiple regions in image ISSN 2224-610X (Paper) ISSN 2225-0603
ii. Feature Extraction: Process to analyze the (Online) Vol.5, No.3, 2015
regions in image [7] O. Ibrahim, H. ElGendy, and A. M. ElShafee
“Towards Speed Detection Camera System for a

ISSN (PRINT): 2393-8374, (ONLINE): 2394-0697, VOLUME-4, ISSUE-11, 2017

27
 INTERNATIONAL JOURNAL OF CURRENT ENGINEERING AND SCIENTIFIC RESEARCH (IJCESR)

RADAR Alternative,” in Proc.11th

International Conf. on ITS Communications,
Sait-Petersburg, Russia, August 2011
[8] A G Rad, A Dehghani and M R Karim,”
Vehicle speed detection in video image
sequences using CVS method”, International
Journal of Physical and Science, Vol. 5(17), pp.
2555-2563, 18 December, 2010
[9] K Shedbalkar, A P Dhamangaonkar and A B
Patil,” Speed estimation using extended Kalman
filter for PMSM”, IEEE conference on Emerging
Trends in Electrical Engineering and Energy
Management (ICETEEEM), 2012,pp-433 to 435
[10] A V Leite, R E Araujo and D Freitas ,” A
new approach for speed estimation in induction
motor drives based on a reduced-order extended
Kalman filter”, IEEE International Symposium
on Industrial Electronics, vol-2, 2004, pp-1221-
1226
[11] N Kassem, A. E. Kosba and M Youssef,”
RF-Based Vehicle Detection and Speed
Estimation “, IEEE 75th conference on
Vehicular Technology, pp-1-5.

ISSN (PRINT): 2393-8374, (ONLINE): 2394-0697, VOLUME-4, ISSUE-11, 2017

28
View publication stats