Journal of Theoretical and Applied Information Technology

20th February 2016. Vol.84. No.2

© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

VEHICLE COUNTING AND VEHICLE SPEED

MEASUREMENT BASED ON VIDEO PROCESSING
1
HARDY SANTOSA SUNDORO, 2AGUS HARJOKO
1,2
Department of Computer Science and Electronics
Universitas Gadjah Mada, Indonesia
E-mail: hardy.santosa.s@mail.ugm.ac.id, 2aharjoko@ugm.ac.id
1

ABSTRACT

Vehicle counting system and vehicle speed measurement based on video processing are few of
systems that utilize digital image processing system as a detector of a moving object such as a vehicle to do
the counting and speed measurement. This system is an early stage in the development of Intelligent
Transportation System. The methods used in this system are background subtraction with Gaussian Mixture
Model (GMM) algorithm and blob detection. Background subtraction method is used because it is a
method that can separate foreground and background smoothly and adaptive to the condition of the frame.
The blob detection method provide the coordinates in the form of centroid so it will show the movement of
the vehicle. System trials conducted on three conditions, namely in the morning, afternoon, and evening.
Speed calibration testing parameters obtained with the use of a speed gun. The accuracy of vehicle counting
obtained by the evaluation method system by comparing the real situation with the results of the system.
The value of the vehicle speed measurement uncertainty obtained by using standard deviation calculations
and the combined uncertainty. Vehicle counting accuracy obtained in the morning conditions is 75.69%.
Vehicle counting accuracy obatined in the afternoon is 90.50%. Accuracy of counting vehicles obtained
85.31% in the evening. The value of the uncertainty around those conditions is approximately ± 3 km/h.
Keywords: Intelligent Transportation System, Video Processing, Blob Detection, Background Subtraction,
Speed Gun

1. INTRODUCTION Intelligent Transportation System (ITS) is the

development of information and communication
Traffic density becomes very important to know technologies in the transportation system. ITS is
especially in the face of the current number of used to improve and streamline mobility, reducing
vehicles passing through a road. A surveyor or the number of accidents and deaths, as well as
observer is usually placed at the curb for counting preserving the environment.
vehicles passing in front of him. The observer
To achieve these objectives, the following
equipped with a tool such as a counter and then
components are identified as the main function of
press the button counter one by one if a vehicle
ITS, namely [1]:
passed. Implementation of the survey was done
manually will lead to human error in the counting 1. Maintaining road serviceability and safety
process in case the density of the number of passing 2. Traffic control
vehicle, environmental influences or internal 3. Travel and user information
condition of the observer itself, resulting in less 4. Demand management
accurate counting process performed by a surveyor. 5. Network monitoring
Besides vulnerable to human error, the counting
was done by humans require a separate fee for each Vehicle counting system and vehicle speed
measurement are few of systems to achieve the goal
implementation so it became less efficient.
of ITS especially network monitoring. This system
Other than passing vehicle on the road, is an early stage in the development of Intelligent
determination of the traffic density required data on Transportation System. These systmes use video
the speed of a vehicle crossing the road. By processing with background substration method and
knowing the speed of passing vehicles then we can blob detection. These systems can be a solution for
conclude the estimated density of the road. counting vehicle and speed automatically.

233
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

2. RELATED WORK well as calculate and measure the speed of passing

vehicles. The design of the hardware is used to
Chen, et al. [2] conducted a vehicle counting
retrieve the data video of passing vehicles.
using blob analysis. The advantages of this system
3.1 User Requirement
is the accuracy of the calculation reaches 90%. The
only drawback of this research is the data retrieval In the system of counting and measuring vehicle
is only done during the afternoon. speed, the system requires information about road
Rahim, et al. [3] calculates the vehicle speed by conditions, particularly the passing vehicles. The
transforming a 2D image into a 3D image to obtain resulting outputs of the system are the number of
the position of the vehicle in three-dimensional passing vehicles and the vehicle speed of each
space. This method consists of image processing, vehicle are successfully detected. Systems designed
extraction centroid, and tracking. Disadvantages of using hardware and software as follows:
this system is the algorithm to calculate the speed 1. Laptop AMD A6-3400 M APU with
has not been validated with actual speed so that the Radeon™ HD Graphics 1.40 GHz RAM 4096
accuracy of the estimated speed can not be MB.
determined. 2. Webcam Logitech C525.
Nurhadiyatna, et al. [4] calculate the speed of a 3. Camcorder JVC Full HD.
vehicle using CCTV cameras and using the PCA 4. Tripod.
method. Speed calibration use three agents who 5. OpenCV 2.4.3.
drove cars and was provided by GPS. The 6. Speed Gun.
advantages of this system are using CCTV cameras 3.2 Data Video Retrieval Scheme
and PCA method. The disadvantages of this system
is the accuracy results varied, ranging from 63% to In designing the vehicle counting system and
99.5%. measuring vehicle speed, the technology used is a
Sina, et al. [5] developed a research to calculate video image processing via cameras mounted on
the number of vehicles and calculate the speed of the highway bridge to monitor the condition of the
vehicles with headlight detection method and toll road. Video data used for testing later is a video
normalized cross-correlation. The advantages of containing footage of the actual road conditions.
this system is more focused on methods headlight Figure 1 shows the schematic capture video data.
detection by using light from a vehicle that can be
represented as a car.
Pour Mohammad and Dehghani [6] used
background subtraction method with estimation
based on the intensity of the color brightness. By
using this method, moving objects in the image are
sequentially detected and the final velocity of the
moving object is measured based on geometric
calculation. The disadvantage of this system is the
accuracy rate of the vehicle reaches ±10% on the
degree of uncertainty. The other drawback of this Figure 1: Video Data Retrieval System Scheme
experiment is the data retrieval process that was In the process of data retrieval, the webcam is
performed during normal conditions. placed on the highway bridge type of concrete with
In order to make a better research, we use speed a height of ± 5.5 meters. This height is the standard
gun for calibration the speed rather than using height of the highway bridge in Indonesia. Concrete
speedometer. For vehicle counting, system trials highway bridge type was chosen so that when
conducted on three conditions, namely in the vehicles such as trucks and buses pass, webcam
morning, afternoon, and evening so that the results will remain stable.
can be seen in any conditions. Mileage trajectory based on measurement is ± 26
3. SYSTEM DESIGN meters. A surveyor placed at a distance of 6 meters
from the highway bridge. Beside the surveyor is a
The design of the vehicle counting system and camcorder that serves as a vehicle recorder from a
vehicle speed measurement based video processing horizontal position and vehicle speed clarification
includes program design and hardware design using tool from speed gun. Conditions mileage is then
computer vision technology. The design of the used in a computer system to get a distance based
program includes the ability to detect the vehicle as on the pixels unit.

234
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

The 60º angle determination is based on research objects (foreground) and background smoothly and
that has been conducted by Li, et al. [7]. According adaptive to the frame conditions. In general, the
to Li, at 60º angle, the measurement result is more steps being taken are as follows:
accurate and maximum. The calibration used in this 1. Lane masking
experiment to detect and measure the vehicle speed Splitting the part of the road where
is in the form of speed gun that was produced by vehicles are moving in one route.
PT. Jasa Marga Persero (Tbk) Branch Semarang, 2. Background elimination
Indonesia. Speed gun is held by a surveyor beside Separating the image of the vehicle from the
the highway so that when a vehicle passes, the road to eliminate image noise that may appear
surveyor will be able to "shoot" a passing vehicle as shadows, rain, etc.
with a speed gun to obtain vehicle speed. 3. Blob dan noise filtration
Making the boundary box (blob) of a moving
3.3 Logic Design Alghorithms
vehicle on the object and removing a small
The design of software used in this system is part
noise to be around the image of the vehicle.
of image processing system. The input of this
4. Countour labelling
system is a video taken with the camera placed
Marking a vehicle that has been detected.
above the highway. In this research, the designs of
5. Vehicle tracking
the software are divided into four parts, namely the
calibration distance, image preprocessing, vehicle First of all, the image obtained will be converted
detection and vehicle counting also vehicle speed into the format of grayscale. This is done because
measurement. the monitoring systems are created using
algorithms that require grayscale format as input.
3.3.1 Calibration distance After that, to eliminate the noise, the image will be
Calibration distance is an important part of the transformed to blur.
measurement process to obtain accurate result.
Distance calibration process carried out in The road being observed shall be referred to the
accordance with the real conditions on the highway region of interest (ROI). At ROI, if there are
when shooting video data. Distance track is unwanted parts to be detected such as sidewalks
bounded by the ROI (Region of Interest) as a and trees, then they will be removed. In order to
regional area of vehicle counting as well as vehicle make the data process more efficient, as well as to
speed measurement. Path length caught on camera avoid other unwanted objecs, the background
is then measured manually by using a meter wheel elimination process is used.
in meters. After that, the next step is the Blob detection is used to detect the shape of the
determination of the pixels length on the system vehicle. All objects that exist in the region of
based on real conditions. interest will be sought by the edge contours of the
3.3.2 Image preprocessing vehicle. Blob detection will detect and count the
The first stage is taking image using the camera. number of vehicles that pass the counting line.
In this process, the vehicle as the main object Figure 2 shows the flowchart of the vehicle
captured by the camera in the form of a frame. counting system and vehicle speed measurement.
Then preprocessing procedure is repeated by
capturing the frame to produce a video. Video
resolution in the system is displayed at the size of
320 x 240 pixels to optimize the image processing
computation. The use of measuring resolution of
320 x 240 pixels is also done due to computer
limitation.
3.3.3 Vehicle detection
The systems use blob detection method and
background subtraction. The use of this method is
based on the results of previous studies that have
good research final results in accuracy. In addition,
blob detection method is used because the method
makes it easier to detect objects in the vehicle
carefully. Background subtraction method is used Figure 2: Flowchart of the Vehicle Counting System and
because it is a method that can separate moving Vehicle Speed Measurement

235
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

The centroid value is obtained from blob (2)

detection method which will be used to get mileage
and tracking of passing vehicles. By using centroid
value, we will know the vehicles’s travel time, by For:
substraction the starting time when centroid appears Original track = actual path length (meters)
from the final one. Track image = path length in the image (pixels)
K factor = multiplier based on the results of
The process for detecting the vehicle speed the speed gun calibration
started with finding the displacement distance
between pixels. Displacement between pixels is K factor is determined by dividing the system’s
determined by the coordinates of the center point of result with the speed gun’s results. In this study, k
factor value is obtained by taking 10 results then
the frame through the coordinates and the
compare it with the speed value from the speed
coordinates. Displacement distance will only be
detected if the object moves into the Region of gun.
Interest (ROI). At the end of the ROI is determined 3.4. Region of Interest
counting line, so that when the object moves pass
counting line then the system will start the process Before going to the counting process of the
of calculating and measuring the speed of vehicles. vehicle and the vehicle speed measurement, in
The distance calculated by the Euclidean distance advance, we have to determine the area of ROI
formula as in Equation (1). Euclidean distance (Region of Interest). ROI is the area that became
formula is used to detect the position of the the observation area of moving vehicle. In the
displacement between pixels. The formula for system, the ROI area is at the y coordinate pixel
calculating the Euclidean distance is: point 0 to 140 pixels, while the y coordinate pixel
point is between 10 to 200 pixels. The
determination of ROI area should be adapted to the
(1) conditions when shooting the video.

For:
= x coordinates of the end point in the region of
interest (pixels)
= x coordinates of the starting point in the region
of interest (pixels)
= y coordinates of the end point in the region of
interest (pixels)
= y coordinates of the starting point in the
region of interest (pixels)
Once the distance trajectory of the vehicle is
known, then the next process is to determine the
travel time of the detected vehicle. In this system,
the processing time is when a vehicle is detected in
the beginning as a blob and ends at the counting Figure 3: Region of Interest
line of the indicator region of interest.
Figure 3 shows the length of the y-coordinates
3.3.4 Calculation to measure the speed of vehicle that is 140 pixels with the actual distance is ± 20
In this system, the travel time unit meters. The length of the x coordinates in the
measurement used is microseconds, because the system is 190 pixels with the actual distance is ±
displacement time between the frames are very 3.5 meters. Once ROI area is defined, then the
quick. After obtaining the object displacement vehicle counting and measuring vehicle speed
distance of the vehicle and its difference in time followed by the detection of vehicle.
then the image speed value can be found by Figure 3 shows that above the counting line there
Equation (2). The formula for calculating the image is condition B. Condition B means the Region of
speed is: Interest (ROI) before the counting line. As for,
below the counting line, there is a condition A. The
condition A is the Region of Interest (ROI) after

236
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

passing the counting line. The logic used in this Table 1: Evaluation method based on test data and
experiment is when a vehicle is in the state B and predictions (Powers, 2007)
then passes the counting line, the system will Real Conditions
calculate and measure the speed of vehicles. After
the object passes the counting line and in condition
A, the system stops the vehicle detection as well as
the calculation and measurement of vehicle’s True False
speed. System Results True TP FP
3.5 Calibration Process Using the Speed Gun False FN TN
In this study, we use speed gun as the calibrator
for speed that obtained in our systems. The k factor
is determined by results of the speed gun. Result of the comparison gives four
In the speed measurement calibration used a possibilities, namely:
calibration tool in the form of speed gun owned by
PT. Jasa Marga Persero (Tbk) Semarang Branch. • True Positives (TP) = A vehicle is detected in
The Speed Gun is derived from the speed real conditions and declared as a vehicle on the
calibration company called Laser Technology of systems.
the United States. Sightings of the speed gun, • False Positives (FP) = None of vehicle is
shown in Figure 4. Speed Gun is newly owned by detected in real conditions but declared as a
PT. Jasa Marga Persero (PT) Semarang Branch in vehicle on the systems.
2014, so that the written calibration date is March • True Negatives (TN) = None of vehicle is
23, 2014. The speed gun has a speed accuracy of ± detected in real conditions and declared as not a
2 km/h. vehicle on the systems.
• False Negative (FN) = A vehicle is detected in
real conditions but declared as not a vehicle on
the systems.
Accuracy is the level of prediction that is
closest to the actual label. Higher values of
accuracy indicates better performance. The truth
value of this parameter is calculated from the
results of True Positive (TP) and True Negative
(TN) divided by the combined results of the
evaluation system consisting of True Positive (TP),
True Negative (TN), False Positive (FP), and False
Negative (FN). Evaluation for accuracy is
represented by Formula 3.

(3)

Figure 4: Speed Gun

4.2. Vehicle Speed Measurement System
4. System Evaluation Method
In the vehicle speed measurement system,
4.1. Vehicle Counting System evaluation system is determined by analyzing the
value of uncertainty which is obtained from the
The evaluation measurements to calculate the difference between the system’s speeds minus by
number of vehicle is done by the method of the average difference, then divided by the amount
prediction based data. Table 1 shows the conditions of data retrieved, minus one. The formula to
based on the comparison between the results of calculate the standard deviation is down below [8]:
prediction systems with real conditions.

237
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

(4) 5.1. Testing In the Morning

It was still dark during the video taping, lux
meter showed a value of 19 lux at an early time
For: data retrieval and 118 lux at the end of data
= standar deviation retrieval. Table 2 shows the calculation results of
= Value of the difference between the vehicle.
measurement results and the speed gun Table 2: Vehicles Calculation Results in the morning
= The average value of measurement result No Conditions Amount Unit Total
= Amount of data 1 The number of 127 1 127
passing vehicles in
After the standard deviation is obtained, it is the real conditions
then being followed by calculating measurement 2 The number of 126
uncertainty. Measurement uncertainty is defined as vehicles counted in
a parameter associated with the measurement the systems
results which are expressed as a reasonable Counted as one 94 1 94
distribution of value that can be assigned to the Counted as two 15 2 30
measurand. If the estimated value of the measurand Distraction 2 1 2
Uncounted vehicles 18 1 18
expressed by and the uncertainty of
measurement, to a certain confidence level, Sometimes one vehicle is detected as two due
expressed by , then the value of the measurand to the number of blob emergence. The cause of the
proficiency level is believed to be in the range emergence of more blob is due to the influence of
according to the formula below: spotlights from the vehicles. Spotlight vehicle is
detected as a foreground, so that the system reads
(5) the spotlight as a blob. Figure 5 shows the creation
of a vehicle that is counted as two vehicles.
In a measurement process, the uncertainty is
estimated from observations of samples of the
measurand ( ). Measuring the amount of n
samples, standard uncertainty can be calculated
using Equation 6 [9].

(6)

For:
= Uncertainty
= Standard deviation
= Amount of data

5. RESULT AND DISCUSSION

Figure 5: A Vehicle is counted as two vehicles
To determine the performance of the system
The accuracy is obtained by using evaluation
made, testing needs to be done in order to take
system based on data test.
some of the original data in the form of video that
records from the actual condition of the highway
with the following specifications:
1. The duration of each video is ± 15 minutes.
2. The webcam has a frame rate of 30 fps.
3. The video resolution is 320 x 240 pixel.
4. The number of vehicles being shot by a speed
gun should be at least 30 vehicles. This number
is needed in order to analyze its uncertainty.

238
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

The results of vehicle’s speed measurement 5.2 Testing In the Afternoon

results in the morning condition are presented in
Sufficient sunlight blazing includes data capture
Table 3.
video during the afernoon, lux meter showed a
Table 3: Vehicle speed measurement results in the value of 1598 lux at an early time data retrieval and
morning 1922 lux at the end of data retrieval. Table 4 shows
Speed the calculation results of vehicle.
Shooting
Speed From
Difference Range Table 4: Vehicles Calculation Results in the afternoon
Vehicle Program Speed
(km/h) Speed
(km/h) gun No Conditions Amount Unit Total
gun (m)
(km/h)
1 The number of
1 88 51 -37 19.0
2 60 48 -12 25.3 passing vehicles in 155 1 155
3 62 57 -5 16.2 the real conditions
4 59 40 -19 34.3 2 The number of
5 90 63 -27 27.6 vehicles counted in 146
6 59 47 -12 24.2 the systems
7 60 46 -14 20.4 Counted as one 140 1 140
8 103 96 -7 27.3 Counted as two 3 2 6
9 61 78 17 23.4 Uncounted vehicles 12 1 12
10 59 54 -5 30.9
11 62 72 10 27.8 The accuracy is obtained by using evaluation
12 84 97 13 40.0 system based on data test.
13 47 67 20 32.6
14 60 66 6 38.1
15 60 93 33 29.7
16 62 77 15 29.5
17 63 53 -10 23.3
18 66 74 8 30.8
19 60 65 5 25.2
20 47 58 11 27.7
21 59 72 13 29.8
22 62 66 4 31.2
23 60 86 26 39.1
24 61 72 11 21.2
25 92 42 -50 39.1 During daylight condition, best value obtained is
26 88 82 -6 28.6 90.5%. This is due to the absence of shadows that
27 89 72 -17 20.7 can cause noise in detecting the real number of
28 88 39 -49 29.5
vehicles. In this condition, it appears that the
29 58 90 32 39.4
30 61 77 16 45.2 shadow falls just below the car and not widen.
31 59 26 -33 18.6
The results of vehicle’s speed measurement
32 97 53 -44 19.1
33 58 79 21 18.8 results during daylight condition are presented in
34 61 98 37 34.3 Table 5.
35 113 92 -21 33.6
36 59 79 20 39.4 Table 5: Vehicle speed measurement results in the
37 60 62 2 28.6 afternoon
38 93 81 -12 23.8
39 60 56 -4 38.2 Speed
Shooting
40 64 56 -8 40.8 Speed From
Difference Range of
Vehicle Program Speed
41 88 80 -8 39.7 (km/h) Speed
(km/h) gun
gun (m)
The number of vehicles that were shot by (km/h)
1 63 86 23 32.2
speed gun then being analyzed to obtain standard 2 90 71 -19 30.7
deviation and uncertainty value. Based on the 3 95 44 -51 32.1
analysis, the standard deviation obtained is 21.77 4 65 79 14 25.8
km/h. While the value obtained for the uncertainty 5 63 74 11 43.3
is ± 3 km/h. 6 59 38 -21 20.8
7 59 68 9 22.8
8 88 36 -52 20.7
9 88 72 -16 35.1
10 90 74 -16 28.8
11 60 63 3 32.4
Vehicle Speed Speed Difference Shooting

239
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Program From (km/h) Range of Table 6: Vehicles Calculation Result in the afternoon
(km/h) Speed Speed towards evening
gun gun (m)
(km/h) No Conditions Amount Unit Total
12 60 66 6 35.9 1 The number of
13 88 40 -48 16.1 passing vehicles in 240 1 240
14 58 68 10 36.4
the real conditions
15 44 78 34 29.7
16 60 51 -9 22.2 2 The number of
17 60 89 29 24.6 vehicles counted in 227
18 92 57 -35 16.3 the systems
19 60 55 -5 19.2 Counted as one 206 1 206
20 59 91 32 23.7 Counted as two 6 2 12
21 90 59 -31 27.6 Counted as three 3 3 9
22 62 64 2 28.4 Uncounted vehicles 25 1 25
23 49 73 24 21.2
24 60 52 -8 20.0 The accuracy is obtained by using evaluation
25 60 54 -6 16.1 system based on data test.
26 92 50 -42 37.1
27 60 80 20 27.9
28 64 107 43 27.8
29 62 73 11 31.5
30 60 70 10 35.2
31 61 80 19 38.6
32 62 61 -1 28.4
33 91 85 -6 45.3
34 60 86 26 42.0
35 101 70 -31 38.3
36 59 84 25 22.2
37 65 56 -9 29.3
38 59 89 30 27.1 The majority of errors are caused by a lot of
39 60 56 -4 21.2 uncounted vehicles, due to poor lightings
40 61 74 13 32.1 conditions during the night. In the dark, the system
41 58 61 3 16.5
could not distinguish between the background and
42 59 40 -19 35.3
43 91 75 -16 26.3 the moving object (foreground). Moreover,
44 59 45 -14 26.1 tollways are not always equipped with adequate
45 58 54 -4 25.4 lightings.
46 90 64 -26 32.0
47 89 79 -10 28.6 The vehicle speed measurement’s results in the
48 60 51 -9 29.4 afternoon towards evening are presented in Table 7.
49 63 81 18 33.2
50 61 60 -1 31.9
51 61 77 16 64.9
Table 7: Vehicle speed measurement results in the
52 60 51 -9 40.8
afternoon towards evening
The number of vehicles that were shot by speed
gun then being analyzed to obtain standard Speed
Shooting
Speed From
deviation and uncertainty value. Based on the Vehicle Program Speed
Difference Range of
analysis, the standard deviation obtained is 22.90 (km/h) Speed
(km/h) gun
gun (m)
km/h. While the value obtained for the uncertainty (km/h)
is ± 3 km/h. 1 64 71 7 33.1
2 45 24 -21 16.9
5.3. Testing in the afternoon towards evening 3 60 76 16 34.7
4 67 93 26 27.9
Lux meter showed a value of 331 lux at an early 5 61 71 10 24.9
time data retrieval and 0 lux at the end of data 6 59 120 61 32.4
retrieval. Table 6 shows the results of the vehicle 7 59 71 12 23.4
calculation in the afternoon towards evening. 8 66 65 -1 17.0
9 63 64 1 21.8
10 61 53 -8 41.6
11 92 61 -31 24.8
12 93 72 -21 26.9
13 63 67 4 40.8
14 63 80 17 38.1

240
 Journal of Theoretical and Applied Information Technology
20th February 2016. Vol.84. No.2
© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Speed
Shooting when passing on the highway. The uncertainty
Speed From value for vehicle speed measurement is ± 3 km/h
Difference Range of
Vehicle Program Speed
(km/h) gun
(km/h) Speed for all conditions.
gun (m)
(km/h)
15 59 92 33 24.7
To minimize the disruption in the form of
16 60 71 11 37.4 shadow, a shadow remover program can be used or
17 59 71 12 39.3 by enlarging the pixel size of the blob.
18 59 37 -22 28.0
19 59 44 -15 33.5 REFERENCES:
20 62 85 23 39.1
21 45 76 31 45.5
22 61 83 22 37.5 [1] AIPCR, Road Network Operations Handbook,
23 63 77 14 20.2 World Road Association, 2003.
24 64 86 22 42.5 [2] Thou-Ho Chen; Yu-Feng Lin; Tsong-Yi Chen,
25 59 78 19 44.4
26 65 124 59 33.8 "Intelligent Vehicle Counting Method Based
27 59 80 21 30.8 on Blob Analysis in Traffic Surveillance,"
28 59 62 3 32.3 Second International Conference on
29 60 67 7 37.4 Innovative Computing, Information and
30 62 59 -3 31.3 Control, 5-7 Sept. 2007, pp.238-238
31 58 56 -2 24.8
32 60 83 23 52.8
[3] Rahim, H.A.; Ahmad, R.B.; Zain, A.S.M.;
33 65 70 5 25.7 Sheikh, U.U., "An adapted point based
34 59 48 -11 24.4 tracking for vehicle speed estimation in linear
35 59 65 6 36.6 spacing," 2010 International Conferenece on
36 63 81 18 19.2 Computer and Communication Engineering
37 61 55 -6 24.0
(ICCCE), 11-12 May 2010 , pp.1-4.
38 61 93 32 38.0
39 88 59 -29 40.0 [4] Nurhadiyatna, A.; Hardjono, B.; Wibisono, A.;
40 62 62 0 27.0 Jatmiko, W.; Mursanto, P., "ITS information
41 61 72 11 29.7 source: Vehicle speed measurement using
42 59 48 -11 18.5 camera as sensor," 2012 International
43 58 64 6 33.2 Conference on Advanced Computer Science
44 59 79 20 44.3
45 36 70 34 33.6
and Information Systems (ICACSIS), 1-2 Dec.
46 62 85 23 42.7 2012, pp.179-184.
47 62 77 15 46.2 [5] Sina, I.; Wibisono, A.; Nurhadiyatna, A.;
Hardjono, B.; Jatmiko, W.; Mursanto, P.,
The number of vehicles that were shot by speed "Vehicle counting and speed measurement
gun then being analyzed to obtain standard using headlight detection," 2013 International
deviation and uncertainty value. Based on the Conference on Advanced Computer Science
analysis results, the standard deviation obtained is and Information Systems (ICACSIS), 28-29
19.55 km/h. While the value obtained for the Sept. 2013, pp.149-154.
uncertainty is ± 3 km/h. [6] Dehghani, A.; Pourmohammad, A., "Single
6. CONCLUSION & FUTURE WORKS camera vehicles speed measurement," 2013
8th Iranian Conference on Machine Vision
In this study, our systems can be used to count and Image Processing (MVIP), 10-12 Sept.
vehicle and its speed automatically in the highway. 2013, pp.190-193.
These systems have a better result in speed [7] Powers, D.M.W., 2007. Evaluation: From
uncertainty because of the use of speed gun. These Precison, Recall, and F-Factor to ROC,
systems also have a good accuration in vehicle Informedness, Markedness & Correlation,
counting on three conditions namely in the Adelaide.
morning, afternoon, and evening. [8] Standardization, I.O. for, 2015. GUM - English
Based on the research results, the calculation - 4. Evaluating standard uncertainty. Available
process of the vehicle in the morning obtained at: www.iso.org [Accessed 17 September
75.69 % in the accuracy value, while in the 2015].
afternoon obtaining 90.50 %. Finally, in the [9] Standardization, I.O. for, 2006. ISO 3534-1,
evening, the accuracy value reached the percentage Geneva. Available at: www.iso.org [Accessed
of 85.31 %. The measurement accuracy is also 17 September 2015]
influenced by the shadow created below the vehicle

241