AL-Mansour Journal / No.

14/ Special Issue /( Part Two) 2010 ( ‫ ) اﻟﺠﺰء اﻟﺜﺎﻧﻲ‬/ ‫ ﺧﺎص‬/14‫ ﻋﺪد‬/ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر‬

Estimating vehicle speed using image processing

Ammar Awni Abbass

University of Baghdad

Abstract:

This research uses image processing to estimate car speed. Given a sequence
of real-time video of traffic images. The images are converted to monochrome
images then edges are extracted and then quantified the resulting images to
classify the objects and find the cars, after detecting the cars, they can be
tracked in different frames and their speed can be estimated by calculating the
position of the car object in different frames and compare it to a ground truth.
This research uses a different approach to estimate speed, rather than using a
reference object to estimate speed, the Matlab function is used and different
relationships between the objects that are extracted from the image itself. This
method does not need camera calibration and does not need any extra
equipments other than a video camera attached to a computer and can be used
by traffic police to monitor highways car or normal roads, and can be
combined with a system to detect the license plate number to create a
complete system to issue an automatic speed ticket for drivers who exceeds
the speed limits. The system is tested in different roads and the results show
the validity of the used method.

10th Scientific Conference 24-25 Oct.2009 - 127 - 2009 ‫ ﺗﺸﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ‬
 Ammar A. Abbas ‫ ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬.‫م‬.‫م‬

1 Introduction

Image processing has been widely applied to traffic analysis in the last decade
for a variety of purposes. The field of traffic research is very wide and it has
many goals that include queue detection, incident detection, vehicle
classification, and vehicle counting, One of the most popular of these
purposes is to attempt to estimate the speed of a moving car, a vehicle system
measures the vehicle speed, and combined with a system to capture and
recognize the image containing the license plate number of the over speeding
vehicles, it can issue an automatic speeding ticket. In this research, we
address the problem of estimating the vehicle speed using a very simple
hardware (a laptop computer with a camera that does not need to be calibrated)
and a very concise and effective software written using Matlab programming
language is used. A new approach is presented for extracting vehicular speed
information, given a sequence of real-time traffic images that is taken from a
laptop camera that is put in the side of a two sided road to capture the images
of that road when the vehicle pass through as shown in figure (1), the images
are then processed by the program.

Figure1: A typical scene of the proposed method.

The structure of this paper presents the related work in literature in section 2,
followed by the basic idea for the problem in section 3, the system layout
(system used to solve this problem) is in section 4, and the implementation of
the system with a breakdown and description of each of the elements in that
system is in sections 5, the results are provided showing the effectiveness of
the proposed method in section 6, the discussions of the results in sections 7,
the graphical user interface is in section 8, the conclusions are found in
section 9, the future work is found at the end of the document in section 10.

10th Scientific Conference 24-25 Oct.2009 - 128 - 2009 ‫ ﺗﺸــﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤـــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌــﺎﺷـﺮ‬
AL-Mansour Journal / No.14/ Special Issue /( Part Two) 2010 ( ‫ ) اﻟﺠﺰء اﻟﺜﺎﻧﻲ‬/ ‫ ﺧﺎص‬/14‫ ﻋﺪد‬/ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر‬

1.1 Problem Definition

The task here is to automatically estimate vehicle speed from video sequences,
acquired from a laptop camera from a road side. Assuming that the studied
road segments are planar and straight. Vehicles are automatically detected and
tracked along frames. No reference object is placed in the street to guide the
program in the process of detecting the vehicles or their speed.

In this work, a 160x120 indexed images at a frame rate of 29 frames per second
(fps) were used. These are demonstrated to be adequate for reliable analysis,
as well as being small enough to allow efficient processing.

1.2 Motivation

As traffic speed monitoring and vehicle counting using buried induction loops
involves relatively high installation and maintenance costs, highway
authorities switch to noninvasive technologies such as microwave radar,
infrared, ultrasound,

And video detection. The advantages of non-optical techniques are their

relative independence from weather and light conditions, however, most of
these devices need to be installed centered above the single lane they service.
The task here is to find a cost effective method to detect the vehicles’ speed
that is cost effective with moderate error rate and low response time.

1.3 Limitations

Advanced video detection is capable of monitoring several lanes

simultaneously from a side mount position and delivers an image of the traffic
situation, but performance suffers from poor lighting or bad weather
conditions. Furthermore, Real-time video detection is computationally
expensive and relies on high-performance signal processing hardware, large
amounts of memory, and high-bandwidth data links.

The density of installation along a highway route is limited and may not reach
the density necessary to acquire enough data for reliable traffic flow
prediction.

10th Scientific Conference 24-25 Oct.2009 - 129 - 2009 ‫ ﺗﺸﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ‬
 Ammar A. Abbas ‫ ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬.‫م‬.‫م‬

2 Related Work
The literature on estimating vehicles’ speed falls in several categories. Many
papers analyze low frame-rate video taken from an un-calibrated camera and
use it to estimate mean traffic speed (3), using this approach takes advantage
of a known relationship between traffic speed and traffic density to make
tracking of individual vehicles unnecessary additionally, the traffic queue
length measurement using an Image Processing Sensor has occupied a great
space in the related work, they have developed new measurement algorithms
and a new hardware, and they have succeeded in the development of a new
image sensor (4).

Two references is particularly relevant to our work. The papers (1,2) discusses
the development of an unconventional, yet an effective method for estimating
vehicle speed, it presents an algorithm that allows automatic 1D measurements
and subsequent estimation of vehicle speed form single un-calibrated images.
It requires minimal external data (one known distance), if basic assumptions
are adopted.

3 Basic ideas
3.1 Vehicle Tracking:

After recording the video from the camera, the video image sequence is stored
in the Matlab workspace; two simple techniques are used for tracking the
objects (vehicles): frame differencing and background subtraction. The
functions in the Image Processing Toolbox are used. The absolute difference
between successive frames can be used to divide an image frame into changed
and unchanged regions. Since only the vehicle moves, the changed region is
expected to be associated only with the vehicle, or possibly with its shadow.

To begin, each frame is converted to an indexed image, then to a monochrome

image, the unique size of the vehicle and the fact that the vehicle moves and
the background doesn’t move enable efficient vehicle tracking.

3.2 Background Subtraction

Another approach to tracking the vehicle is to estimate the background image

and subtract it from each frame. The approach here is to find the pixel-wise
maximum among several neighboring frames.

10th Scientific Conference 24-25 Oct.2009 - 130 - 2009 ‫ ﺗﺸــﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤـــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌــﺎﺷـﺮ‬
AL-Mansour Journal / No.14/ Special Issue /( Part Two) 2010 ( ‫ ) اﻟﺠﺰء اﻟﺜﺎﻧﻲ‬/ ‫ ﺧﺎص‬/14‫ ﻋﺪد‬/ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر‬

That’s exactly what morphological dilation does, if a structuring element

oriented is used along the frame dimension. Next, the absolute difference
between each frame and its corresponding background estimate is computed.

4 System Layout

The system model can be broken into three primary stages:

1- Preprocessing.

2- Data Reduction.

3- Feature analysis.

In preprocessing, the operation is performed to make data reduction and

analysis easier, the video stream is converted to an indexed image then to a
binary image. The binary (Black and White) image that contains all the
information necessary to discern the object's outline.

In Data Reduction stage the frame differencing is made by using a special

Matlab function. The small objects are removed using the function bwareaopen
this process is made to eliminate any unnecessary objects. Each object in the
image is marked using the regionprops Matlab function and the objects are
properties are measured using the same function.

In the feature analysis, the features of the objects which are extracted by data
reduction process are examined and evaluated for their use in the application.

This stage includes finding the car object in the frames and tracking it through
the different frames and calculating the speed of the vehicles.

The recording of the video operation is not included in the model because, it is
not part of the computer vision process and it is considered as part of the
hardware.

5 Proposed System
The steps of the process are illustrated here with some discussions of each
step. The user must make sure that the image processing tool box is installed
by typing the command (ver), which will display information about the version
of Matlab that we are using, and the toolboxes that are installed in the program.

10th Scientific Conference 24-25 Oct.2009 - 131 - 2009 ‫ ﺗﺸﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ‬
 Ammar A. Abbas ‫ ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬.‫م‬.‫م‬

5-1 Read the Video

After recording the video in the Matlab workspace, the video must be read
using (aviread) function which read the video from Matlab workspace. The past
function doesn’t display the video in any form; its only work is to prepare the
video for the next steps.

5-2 Convert the Video Frame to an Image

The function frame2im converts the single movie frame into an indexed image
and an associated color map. A simple loop is used to repeat this operation for
the entire movie, the dimension of the loop is decided by the function length
which returns the total number of frames in the movie. This will prepare the
images for further processing.

5-3 Compute the Images Absolute difference

The function imabsdiff subtracts each element in current array (current image)
from the corresponding element in the previous array (previous image) and
returns the absolute difference in the corresponding element of the output
array. The output image has the same class and size as current and previous
images. As shown in fig (2). The expected output of this process is the vehicle
object only.

Figure (2): Compute the Images Absolute difference.

5-4 Convert the image to Black and white

The image is converted to black and white by using the function im2bw, which
converts an image to binary image, based on threshold, the threshold is
computed based on a function named graythresh automatically which
computes a global threshold. A threshold that can be used to convert an
intensity image to a binary image combined with im2bw. Each pixel takes a
value that is a normalized intensity value that lies in the range [0, 1].The
graythresh function uses Otsu's method, which chooses the threshold to
minimize the intraclass variance of the black andWhite pixels.

10th Scientific Conference 24-25 Oct.2009 - 132 - 2009 ‫ ﺗﺸــﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤـــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌــﺎﺷـﺮ‬
AL-Mansour Journal / No.14/ Special Issue /( Part Two) 2010 ( ‫ ) اﻟﺠﺰء اﻟﺜﺎﻧﻲ‬/ ‫ ﺧﺎص‬/14‫ ﻋﺪد‬/ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر‬

The results are shown in fig (3).

Figure (3): Convert the image to Black and white

5-5 Label the objects in the Image and Find the Center of the vehicles to keep
track of them:

All the components in the image must be marked, By doing this the number of
labeled objects is found in (numobjectsbase) parameter, this step is related
with another step to measure the all the properties of the image components
such as the area of the components and the center of each components.

Since the car is the largest object in the image if we call the (centroid) property
of the (regionprops) function, we can obtain the center of the vehicle object.

5-6 Convex Hull Extraction

After the converting to black and white above, moving edges are filled and
appear as solid moving blobs. To characterize the blobs, we use a convex hull
to approximate the contour of the vehicles. In many cases, a convex hull is a
good approximation of the projection of a car. We call the (ConvexHull)
property of the (regionprops) function.

5-7 Bounding Box Extraction

To obtain scaling information directly from the image rather than using explicit

camera calibration, we exploit the known geometric relationships in the

images. We do this by constructing a bounding box to enclose the convex hull.
This bounding box is used to isolate the area of interest in the image and is
similar to window (or key region) processing. This can be easily made by
calling the (Bounding Box) property of the (regionprops) function.

5-8 Geometric analysis

The direction of motion of each vehicle is fixed, the best fit line is computed
through the centroids of the convex hulls found in a series of images .

10th Scientific Conference 24-25 Oct.2009 - 133 - 2009 ‫ ﺗﺸﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ‬
 Ammar A. Abbas ‫ ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬.‫م‬.‫م‬

Ground truth distance is estimated by using scale information along the

direction of motion, and these distances, with the frame rate of the video
sequence, are used to estimate speed. The function bwdist computes the
distance between the centers of the convex hull.

To get the scale information from the images directly, the Matlab function
bwdist computes the Euclidean distance transform of the binary image directly.
The Euclidean distance between (x1,y1) and (x2,y2) is :

……….. (1)

The first point (x1,y1) and the second point (x2,y2) are the centers of the
vehicle object in first and last points of a line that passes through the centers
of the vehicle object in the different frames, the centers is found using the
centroid property of the (regionprops) function. When the vehicle travels a
predefined fixed distance the number of frames that is necessary to travel that
distance in the image is calculated and it is compared to a ground truth which
is the number of frames that is necessary to vehicle to travel the same distance
in order to achieve certain speed.

For example if the number of frames that is needed to travel the fixed distance
in a certain speed is 80 frames, we can estimate the speed from the following
table (1) which presents a pre-calculated number of frames at a different
speeds:

Speed(Km/h) No. of frames

1 25 110

2 30 98

3 35 87

4 40 77

5 45 65

6 50 54

Table (1): A pre-calculated speed versus number of frames

(Note: This table is a part of a larger table that gives the speed versus the
number of frames for each speed from (25-130) Km/h)

Each of the parameter in table(1) is entered in a separate 1D matrix, named

Speed and Frames and each new speed can be estimated by calculating the

10th Scientific Conference 24-25 Oct.2009 - 134 - 2009 ‫ ﺗﺸــﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤـــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌــﺎﺷـﺮ‬
AL-Mansour Journal / No.14/ Special Issue /( Part Two) 2010 ( ‫ ) اﻟﺠﺰء اﻟﺜﺎﻧﻲ‬/ ‫ ﺧﺎص‬/14‫ ﻋﺪد‬/ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر‬

number of frames can be used to get the speed by using interpolation, the
Matlab function (interp1) can be used to achieve this process. And for a car
that travel the fixed distance in 80 frames the estimated distance using
interpolation is expected to be 38.5 Km/h.

If the work is made manually by programming and not by the Matlab builtin
functions, the triangular relationship within the bounding box must be
exploited. The pixel length, L_pixel, of a vehicle is estimated along the best fit
line (L) indicating the direction of travel. It is estimated to be the length of the
cord along the best fit line that intersects the bounding box.

L_pixel=box_width/sin ……….. (2)

Where the direction of motion and in this research is equal to 90 degrees, so

that the pixel length is equal to bounding box width. 

To map the pixel length to ground truth vehicle lengths, The empirical vehicle
length distribution is used. This allows the user to obtain the ratio of the
physical length, L_physical, and the pixel length, L_pixel, which is the scale
factor s:

S= L_physical/ L_pixel(ft/pixel)…..(3)

The travel distance between frames and the vehicle speed is estimated using
the scale factor just obtained from the above geometric analysis. Assuming
that distance traveled by a vehicle is defined by the displacement of its
centroid.

To obtain the physical distance of moving centers, the scale factors at each
pixel along the travel path having angle of is estimated. To this end, the total
number of pixels along the travel path needs to be estimated, which can be
obtained by using:

#of pixels along moving angle =#of vertical pixels/sin  

Where the number of vertical pixels is simply the vertical pixel length between
the first and the last centroids.

10th Scientific Conference 24-25 Oct.2009 - 135 - 2009 ‫ ﺗﺸﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ‬
 Ammar A. Abbas ‫ ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬.‫م‬.‫م‬

For an image sequence with k frames, where s1 is the scale factor at the
centroid of the convex hull of the vehicle of interest in the first frame, sk is the
scale factor at the centroid of the convex hull of the vehicle in the k-th frame,
and the number of pixels along the driving Path between these two centroids is
n; the scale change per pixel, s, ca be computed as:

s=s k - s1 /n-1 (ft / pixel2 ) ………..(5)

The total distance D traversed between the images is then obtained by

summing up the scale factor series as:

D= n*((s1+sk)/2)………………..(6)

The speed is then estimated as the ratio of the interframe travel distance and
the known frame rate.

6 Results
Testing of the system will first begin with loading artificial or acquired images
to the software with a certain result expected. Testing will be as simple as a car
traveling in front of the laptop camera and its motion viewed on the screen as
shown in figure (1).

During the testing, 40 different videos are fed to the system (27.5 hours of
video containing in total of 110,918 frames) have been captured, the frame rate
was 29 frame per second, each frame was 160*120 pixel in size. Each video
contains many cars, and each car speed is measured by the program
individually.

The drivers of the vehicles were asked to drive at a certain velocity and the
number of frames at each velocity is taken and it is used to make a lockup
table for vehicle speeds that ranges from (25-130) Km/h that table looks like
table(1) but with much more values, this table is used as reference for each
new value measured by the program for the new recorded video. Some
examples of the program execution is given in table(2).

The automatic approach gave a very satisfactory estimated accuracy in

vehicle speed of about ±3 km/h error rate.

10th Scientific Conference 24-25 Oct.2009 - 136 - 2009 ‫ ﺗﺸــﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤـــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌــﺎﺷـﺮ‬
AL-Mansour Journal / No.14/ Special Issue /( Part Two) 2010 ( ‫ ) اﻟﺠﺰء اﻟﺜﺎﻧﻲ‬/ ‫ ﺧﺎص‬/14‫ ﻋﺪد‬/ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر‬

M ¤ M ¤ M ¤

1 23 +1 8 54 -2 15 91 +1

2 24 +1 9 59 -1 16 94 +3

3 29 +1 10 64 -3 17 101 +2

4 33 -2 11 71 -2 18 106 -3

5 41 +1 12 76 +1 19 111 +1

6 43 -3 13 83 +2 20 117 -2

7 49 -1 14 84 -2 21 119 +2

Table (2): Speed Estimation (Km/h), M: is the manually measured speed. ¤: is the error in the automatic
measurements.

The results of measuring the speed of a vehicle that appears in the video can
be found the Matlab command window.

7 Discussions
An efficient image processing techniques are applied to estimate travel speed
from image sequences of moving vehicles. Simple geometric relations are
obtained directly from the image itself and are used without explicit camera
calibration. Furthermore, the techniques presented are validated against
ground truth by field trials.

It was found that the proposed method is applicable and can provide the
means to estimate car speed that can be used by traffic police in suburban
roads, the approach used is very accurate when compared to manual
measurements (table1), the proposed method provide a cost effective method
with a good response time when compared to the specialized machine for
estimating vehicles speed (like radar). The proposed method cost is limited to
a laptop with a camera and the program, while the radar, for example, can cost
up to five thousand dollars for the single device.

10th Scientific Conference 24-25 Oct.2009 - 137 - 2009 ‫ ﺗﺸﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ‬
 Ammar A. Abbas ‫ ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬.‫م‬.‫م‬

8 Graphical User Interface (GUI) for the Program

The Graphical User Interface is shown in figure (4) below:-

Figure (4): Graphical User Interface

Each one of the buttons when pressed will calculate the speed of the vehicle
associated with that sample, for instance, if we press button named 50 Km
(which means that the vehicle in that sample travel at speed 50 Km (pre-
calculated).

After we select one of the samples the (10 Km-130 Km). Images containing the
image absolute difference and the same image after it is converted to black and
white to illustrate the work of the program, the estimated vehicle speed appear
on the Matlab work space.

9 Conclusions
1-The proposed system does not need a professional operator to work, it has a
simple GUI.

2-The system cannot tolerate errors in the capturing the video, errors like
rotating the camera during the video recording or changing the position of the
camera cannot be corrected by the program.

3-The program assumes that there is one vehicle in the street, and it does not
10th Scientific Conference 24-25 Oct.2009 - 138 - 2009 ‫ ﺗﺸــﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤـــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌــﺎﺷـﺮ‬
AL-Mansour Journal / No.14/ Special Issue /( Part Two) 2010 ( ‫ ) اﻟﺠﺰء اﻟﺜﺎﻧﻲ‬/ ‫ ﺧﺎص‬/14‫ ﻋﺪد‬/ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر‬

change lanes, so the software is not designed to be used in crowded streets or

busy highways, the system best usage is in suburban roads.

4-The system is tested in day-light; no attempt is made to test the system in the
nights.

5- The system is tested in a sunny day and no attempt is made to test the
system in the different weather conditions (rain or dust).

10 Future Work
1- Increase the number of lanes that the system must cover.

2- Change the angle of view for the video to top view.

3-Change the system so that it can operate in nights as well.

4-Increase the number of cars that can be processed at the same time.

10th Scientific Conference 24-25 Oct.2009 - 139 - 2009 ‫ ﺗﺸﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ‬
 Ammar A. Abbas ‫ ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬.‫م‬.‫م‬

11 References
1- Lazaros Grammatikopoulos, George Karras, Elli Petsa, (2005). “Automatic
estimation of vehicle speed from uncalibrated video sequences”. International
symposium on modern technologies, education and professional practice
ingeeodesy and related fields, PP (1-5).

2- D.J. Dailey and L. Li. (1999). “An algorithm to estimate vehicle speed using
uncalibrated camera”, IEEE, VOL, 4, NO.8, PP (3-12).

3- Jared Friedman. (2005). “Finding Mean Traffic Speed in Low Frame-rate

Video” IEEE Trans. Intelligent Transportation Systems Vol, (1), No. 2, pp. (98-
107).

4- Masakatsu Higashikubo(1) Toshio Hinenoya(2) Kouhei Takeuchi(3).2001.

“Traffic Queue Length Measurement Using an Image Processing Sensor”. In
Journal of Further and Higher Education, VOL,25,NO.(2).CarfaxPublishing,
PP(10).

5- Rafael M.crouch.”The Matlab image processing toolbox guide”,USA,Adison-

Wisley, 2004,PP(116-150).

6- Rafael C. Gonzalez and Richard E.woods,"Digital Image Processing",

Adison-Wesley, USA 2002, PP (20-80).

7-John K. Wally. “The Matlab image processing toolbox guide”, 2004.Prentice

hall USA, PP (50-70).`

10th Scientific Conference 24-25 Oct.2009 - 140 - 2009 ‫ ﺗﺸــﺮﯾﻦ اﻻول‬25-24 ‫اﻟﻤﺆﺗﻤـــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌــﺎﺷـﺮ‬
‫)‪AL-Mansour Journal / No.14/ Special Issue /( Part Two‬‬ ‫‪2010‬‬ ‫ﻣﺠﻠﺔ اﻟﻤﻨﺼﻮر ‪ /‬ﻋﺪد‪ /14‬ﺧﺎص ‪ ) /‬اﻟﺠﺰء اﻟﺜﺎﻧﻲ (‬

‫ﻣﻌﺎﻟﺠﺔ اﻟﺼﻮر ﻟﺘﺤﺪﯾﺪ ﺳﺮﻋﺔ اﻟﻤﺮﻛﺒﺎت‬

‫م‪.‬م‪ .‬ﻋﻤﺎر ﻋﻮﻧﻲ ﻋﺒﺎس‬

‫ﺟﺎﻣﻌﺔ ﺑﻐﺪاد‬

‫اﻟﻤﺴﺘﺨﻠﺺ‪:‬‬
‫ﯾﺴﺘﺨﺪم ھﺬا اﻟﺒﺤﺚ ﻣﻌﺎﻟﺠﺔ اﻟﺼﻮر ﻟﻤﺤﺎوﻟﺔ ﺗﺤﺪﯾﺪ ﺳﺮﻋﺔ اﻟﻤﺮﻛﺒﺎت اﻟﺘﻲ ﺗﺘﺤﺮك اﻣﺎ اﻟﻜﺎﻣﯿﺮا اﻟﺘﻲ ﺗﻠﺘﻘﻂ ﺻﻮراً ﻓﺪﯾﻮﯾﺔ‬
‫ﻣﺘﺴﻠﺴﻠﺔ )اﻃﺎرات( ﻟﻠﺸﺎرع واﻟﺴﯿﺎرات اﻟﺘﻲ ﺗﺴﯿﺮ ﻓﯿﺔ وﻓﻲ اﻟﺰﻣﻦ اﻟﺤﻘﯿﻘﻲ‪ .‬ﻧﻘﻮم اوﻻ ﺑﺘﺤﻮﯾﻞ اﻟﺼﻮرة اﻟﻰ ﺻﻮرة‬
‫اﺣﺎدﯾﺔ اﻻﻟﻮان)اﺳﻮد واﺑﯿﺾ( ﺛﻢ ﻧﻘﻮم ﺑﺎﺳﺘﺨﻼص اﻟﺤﺎﻓﺎت وﻧﺼﻨﻒ اﻟﺼﻮر اﻟﻨﺎﺗﺠﺔ ﻟﻨﺤﺪد اﻻﺟﺴﺎم اﻟﻤﻮﺟﻮدة ﻓﻲ‬
‫اﻟﺼﻮرة وﺑﺎﻟﺘﺎﻟﻲ ﻧﺤﺪد اﻟﻤﺮﻛﺒﺎت وﻧﻘﻮم ﺑﻤﻼﺣﻘﺔ ھﺬه اﻟﻤﺮﻛﺒﺎت ﻓﻲ اﻻﻃﺎرات اﻟﻤﺨﺘﻠﻔﺔ وﻧﺴﺘﻄﯿﻊ ﺗﺤﺪﯾﺪ ﺳﺮﻋﺔ اﻟﺴﯿﺎرة‬
‫ﺑﺤﺴﺎب ﻣﻮﻗﻊ ﺟﺴﻢ اﻟﺴﯿﺎرة ﻓﻲ اﻻﻃﺎرات اﻟﻤﺨﺘﻠﻔﺔ‪.‬ﯾﺴﺘﺨﺪم ھﺬا اﻟﺒﺤﺚ ﻃﺮﯾﻘﺔ ﺟﺪﯾﺪة ﻟﺘﺤﺪﯾﺪ اﻟﺴﺮﻋﺔ ﻓﺒﺪﻻ ﻣﻦ اﺳﺘﺨﺪام‬
‫ﺷﻮاﺧﺺ ﻓﻲ اﻟﻄﺮﯾﻖ ﻟﺘﺤﺪﯾﺪ اﻟﺴﺮﻋﺔ ﻓﺎﻧﻨﺎ ﻧﺴﺘﺨﺪم دوال ﺑﺮﻧﺎﻣﺞ اﻟﻤﺎﺗﻼب واﻟﻌﻼﻗﺎت ﺑﯿﻦ اﻻﺟﺴﺎم اﻟﻤﺨﺘﻠﻔﺔ اﻟﻤﺴﺘﺨﻠﺼﺔ‬
‫ﻣﻦ اﻟﺼﻮرة ﻧﻔﺴﮭﺎ‪ .‬ھﺬه اﻟﻄﺮﯾﻘﺔ ﻻﺗﺤﺘﺎج اﻟﻰ ﺿﺒﻂ ﻟﻠﻜﺎﻣﯿﺮا ﻛﻤﺎ ﻓﻲ اﻟﻄﺮق اﻟﻤﺴﺘﺨﺪﻣﺔ ﺣﺎﻟﯿﺎ ﻛﻤﺎ اﻧﮭﺎ ﻻﺗﺤﺘﺎج اﻻ اﻟﻰ‬
‫ﻛﺎﻣﯿﺮا ﺑﺴﯿﻄﺔ ﻣﺮﺗﺒﻄﺔ ﺑﺤﺎﺳﺒﺔ وﯾﻤﻜﻦ اﺳﺘﺨﺪاﻣﮭﺎ ﺑﻮاﺳﻄﺔ ﺷﺮﻃﺔ اﻟﻤﺮور ﻟﻤﺮاﻗﺒﺔ اﻟﺴﺮﻋﺔ ﻓﻲ اﻟﻄﺮق اﻟﺨﺎرﺟﯿﺔ‬
‫واﻟﺪاﺧﻠﯿﺔ ‪,‬ﻛﻤﺎ ﯾﻤﻜﻦ ان ﻧﺪﻣﺠﮭﺎ ﻣﻊ ﻧﻈﺎم ﻟﻘﺮاءة وﺗﺤﺪﯾﺪ ارﻗﺎم اﻟﺴﯿﺎرات ﻟﺘﻜﻮن ﻧﻈﺎﻣﺎ ﻣﺘﻜﺎﻣﻼً ﻻﺻﺪار ﻣﺨﺎﻟﻔﺎت اﻟﺴﺮﻋﺔ‬
‫اﻟﯿﺎً ﻟﻠﺴﻮاق اﻟﺬﯾﻦ ﯾﺘﺠﺎوزون اﻟﺴﺮع اﻟﻤﺤﺪدة‪ .‬ﺗﻢ اﺧﺘﺒﺎر اﻟﻨﻈﺎم ﻓﻲ ﻃﺮق ﻣﺨﺘﻠﻔﺔ واﻛﺪت اﻟﻨﺘﺎﺋﺞ ﻣﺼﺪاﻗﯿﺔ اﻟﻄﺮق‬
‫اﻟﻤﺴﺘﺨﺪﻣﺔ وﻗﺪرة اﻟﻨﻈﺎم ﻋﻠﻰ اﻟﻌﻤﻞ ﻓﻲ ﻣﺨﺘﻠﻒ اﻟﻈﺮوف ‪.‬‬

‫‪10th Scientific Conference 24-25 Oct.2009‬‬ ‫‪- 141 -‬‬ ‫اﻟﻤﺆﺗﻤــﺮ اﻟﻌﻠﻤــﻲ اﻟﻌـﺎﺷــﺮ ‪ 25-24‬ﺗﺸﺮﯾﻦ اﻻول ‪2009‬‬