APPLICATION OF DIGITAL CELLULAR RADIO FOR MOBILE LOCATION ESTIMATION


IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

DETECTION OF TRAFFIC DENSITY WITH IMAGE 

PROCESSING USING PIN HOLE ALGORITHM 

MOCHAMAD ADITYA IRAWANTO, CASI SETIANINGSIH* AND BUDHI IRAWAN 

Computer Engineering, School of Electrical & Engineering, 

 Telkom University, Indonesia  

*Corresponding author: setiacasie@telkomuniversity.ac.id 

(Received: 6th July 2021; Accepted:18th October 2021; Published on-line: 4th January 2022)

ABSTRACT: Intelligent traffic monitors have developed and have become more
attractive in recent years. A detection system in the monitoring traffic system is proposed 

using different algorithms. Pin Hole Algorithm used to detect the car that passes the road 

(the studied area). A fixed camera mounted at a predetermined point is used with known 

height (of the camera), the intensity of the light, and the visibility of the camera. The 

classification process is important to know the traffic congestion status. The traffic 

congestion status will be sent to the server address already provided.  In the congestion 

detection test, results were obtained with an accuracy value of 85% using the 64x64 grid 

division and obtaining good detection results for susceptible light intensity values 

between 5430 and 41379 LUX with an accuracy value of between 60% and 90%.  

ABSTRAK: Sejak beberapa tahun ini, sistem pengawasan trafik pintar telah dibina dan 

terus berkembang luas. Sistem pengesanan dalam sistem trafik pengawasan telah 

dicadangkan menggunakan pelbagai algoritma. Algoritma lubang pin digunakan bagi 

mengesan kereta yang melalui jalan (kawasan kajian). Kamera dipasang tetap pada titik 

tertentu iaitu dengan menyelaras ketinggian kamera, keamatan cahaya, dan 

kebolehlihatan kamera. Proses klasifikasi sangat penting bagi menentukan status 

kesesakan trafik. Status kesesakan trafik akan dihantar ke alamat pelayan yang telah 

disediakan. Nilai ketepatan ujian pengesanan kesesakan yang diperoleh adalah 85% iaitu 

menggunakan pembahagi grid 64x64 dan dapatan kajian menunjukkan pengesanan yang 

baik bagi nilai keamatan cahaya antara 5430 dan 41379 LUX dengan nilai ketepatan 

antara 60% dan 90%. 

KEYWORDS: intelligent traffic density; pin hole algorithm; congestion monitor; car 

detection; traffic details; image processing  

1. INTRODUCTION

Traffic lights with less efficient distribution can cause overcrowding at crossroads and

can cause congestion that causes traffic jams to the detriment of road users [1]. Congestion 

is a condition where there is a stagnant or interrupted flow of vehicles passing through the 

road due to the accumulation of vehicles exceeding the existing traffic capacity [2]. 

Congestion problems are caused by a lack of available facilities and infrastructure 

such as road monitoring systems, road widths, traffic signs. Generally, traffic monitoring 

systems that are running today still use traditional systems. The system used now is to use 

cameras installed at various points, which are monitored manually by the Transportation 

Agency. Overall, this system can work well, detecting congestion in accordance with the 

244



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

rules of the city of Bandung in Indonesia. It's just that congestion validation is still a bit 

challenging because there is no definite data comparison yet.  

Pin Hole algorithm was used to obtain the congestion status for traffic control on the 

toll road section. Before that, the camera is fixed to take images (video) for the studied 

area, pre-processing of the image, contrast limited adaptive histogram equalization, edge 

detection, enhancement, then image segmentation and classification. The presented 

algorithm can work with different city roads.  

2.   RELATED WORK 

To maximize the detection of traffic jams using Image Processing, according to 

previous research that has been done [3], several factors must be considered. The first is 

the camera settings. This setting is required to be able to determine the shooting results. 

The accuracy depends on many factors such as the width of the road, the height of the 

camera installation and the angle of the camera. The second is image segmentation. In 

object detection, before filtering using a certain algorithm, the initial process is to convert 

the data format. The data in MPEG form is converted to JPEG. In the form of JPEG 

format will be searching the edge value or background image, edge value is used to 

distinguish between road and object to be processed [4]. The edge value taking is done 

continuously in order to get perfect results [5]. 

Detection of vehicle objects on a road segment to determine the status of traffic jams 

can use the Pin Hole Algorithm [6]. Pin Hole algorithm works by using an IP camera. The 

result of detection from the camera will be MPEG4 which will be converted back into 

MJPEG to be processed to the next stage [7]. Then the obtained image is filtered using a 

Greyscale filter to determine the background value of the image. The background value is 

used to separate where the object is to be detected and which image is just the background 

of the image [7]. The picture shooting takes place within three minutes, and it is because if 

the image is taken less than three minutes sometimes still has data that is not yet valid and 

not perfect. Then determined the parameters of the results in an emphasis on three 

parameters, namely the first parameter is the size of the image based on terms of 

connection, the speed of data as a data source. The second parameter is the large image 

download and the third parameter is the pixel pin hole size. 

Another method that can be used in vehicle counting in traffic control is Blob analysis 

[8]. The Blob Filter algorithm is able to detect congestion well. The steps required in this 

algorithm are as follows: 

1. Vehicle Segmentation, settings in a location of picture shooting on camera.  

2. Background Updating, define the background of the image by segmenting. 

3. Feature Extraction and Vehicle Classification used to be able to define different 
objects when detecting images.  

4. Vehicle Tracking is used to determine the amount of vehicle current based on the 
distance of each vehicle. 

5. Vehicle Flow Analysis and Counting is used for the final count of the number of 
vehicles and the speed of the coming vehicle on the line.  

The output of this method is the number of vehicles outside the track with the average 

speed of the vehicle so that it can be concluded whether the situation can be said to be 

245



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

stuck or not. This method also has an average accuracy value in three conditions of 90% 

with an estimated error of less than 5 km/hour. 

Detecting traffic jams in the city in real-time with the help of a UAV can be chosen as 

another alternative in analyzing traffic conditions. Traffic data in the form of videos taken 

from drones is then detected using a Deep Learning Algorithm, namely Mask RCNN. 

Several steps must be taken, namely, firstly counting the number of vehicles, and 

estimating the speed of the vehicle. Second, estimate the geometric space for pixels in the 

image to perform vehicle recognition and tracking [9]. 

3.   RESEARCH METHOD 

3.1  Illustrations  

Acquisition Video is the process of taking the recording of toll road conditions taken 

with the *.mp4 video format then saved in a personal computer. The video acquisition is 

taken based on the entrance to the Bandung city with the criteria of having one common 

goal and having multiple branches as an alternative way. The video acquisition was taken 

using GoPro Hero5 Black with 720 pixels video quality and 1280 x 720 pixels resolution 

at 30 fps. The video acquisition is taken at the top of a person crossing the bridge (JPO) 

which is on a 5.1 m highway toll road. An illustration of video acquisition can be seen in 

Fig. 1. 

 

Fig. 1: Illustration of video acquisition. 

3.2  Algorithm Steps  

Traffic congestion detection is achieved by many steps such as image acquisition 

from video acquisition, pre-processing on image, Pin Hole processing to perform vehicle 

calculation on image and result from classification. The process of the suggested algorithm 

can be described as follows: 

3.2.1 Image Acquisition 

The image acquisition process is a step to get a sample image based on video 

recording. Image captured with *.mp4 format and stored in *.png format. Figure 2 shows 

the acquisition of the saved image with the format of *.png given region Region of Interest 

(ROI) to detect congestion. ROI is an area marked as the boundary area of toll roads that 

will be calculated using digital image processing. The original image that has been given 

the ROI region can be seen in Fig. 2. 

246



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

 

Fig. 2: Image acquisition with ROI. 

3.2.2 Pre-Processing 

The pre-processing step is used to perform calculations by processing the image by 

combining filters in the image processing. The process that runs in this stage is like 

capturing the image of the video, changing the grayscale filter, inserting the image into the 

CLAHE filter, determining the border of the image, inserting the image into the low pass 

filter¸ and performing the image segmentation. 

a. Greyscale 

Greyscale is a color in pixels that are still between the range of black and white 

values 0-255 which means 0 is a value in black and 255 is white. Gray degree in 

greyscale can be done by using digital image processing by adjusting the gray 

degree of the image using a threshold value. Thresholding is the method used to be 

able to adjust the number of gray-degree values in digital images [10]. The colored 

image has a matrix value for each pixel in which each pixel has a value of red, 

green, and blue (r, g, b). To be able to change the color image into greyscale and 

set the gray degree using thresholding will be explained as follows 

• Calculate the value of greyscale using the equation 

𝑠 = 
𝑟 ×𝑔 ×𝑏

3
 (1) 

where S is greyscale value 

• Calculate the value of thresholding using the equation 

𝑥 = 
𝑊

𝑏
  (2) 

Where 𝑥 is gray degree values after thresholding, W is gray degree values before 
thresholding, and b is the desired quantization values.   

The result of color image processing being the greyscale image can be seen in Fig. 

3. 

 

Fig. 3: Greyscale image result. 

247



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

b. Contras Limited Adaptive Histogram Equalization (CLAHE)  

  CLAHE is a digital image processing process that improves image contrast by 

increasing contrast in local images by providing a maximum threshold value [11]. 

The CLAHE process functions to divide the local image contrast in a symmetrical 

grid called the area size. The image's division region size can be divided into 

sections such as CR for the corner region, BR for border region¸ IR for the inner 

region [12]. The existence of the division of region sizes due to the division of the 

grid there are different characteristic values for each value of the neighborhood. 

The structure of region size can be seen in Fig. 4. 

 

Fig. 4: Region size CLAHE. 

After dividing the local image into a symmetric grid, the calculation is performed 

by obtaining a new gray level on each local grid using the cumulative function 

distribution of histograms used in each local region (i, j). 

• Calculates the distribution using an equation  

(𝑘) = 
(𝑁 −1)

𝑀
 .∑ ℎ(𝑘);𝑛 = 1,2,3,…. ,𝑁 −1

𝑛

𝑘=0
   (3) 

where M is the number of pixels, N is the greyscale value, and h(k) is the 

histogram on grey value.  

After calculating the distribution of cumulative function, then regional on local 

image give maximal threshold limit on the image using CLAHE. 

• Calculates threshold limits using equations 

  𝛽 = 
𝑀

𝑁
(1+

𝑎

100
(𝑆𝑚𝑎𝑥 −1) (4) 

Where M is wide of region size, N is greyscale value, and 𝑎 is minimum and 
maximum limit values of histograms ranging from 0-100. 

The result of image processing into CLAHE filter can be seen in Fig. 5. 

 

 

Fig. 5: CLAHE image result. 

248



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

c. Edge Detection Using Canny Method 

Edge Detection or abbreviated as ED is an object detection process in digital image 

processing that produces an edge output of the image object used to mark the detail 

part of the image [13]. ED is processed using Canny method. The Canny method is 

used because it produces a visible edge shape compared to Robert, Sobel, Prewitt, 

and LoG methods after being processed using the Global Thresholding method 

[13]. ED in the process is done by giving gaussian filter greyscale image which 

then convoluted by using method of Canny to get result of fine edge image. 

Gaussian filter is done by using gaussian filter equation as follows [14]: 

• Calculate Gaussian filters using equations 

𝐺(𝑥) = 
1

√2𝜋𝜎2
𝑒
− 
𝑥2+ 𝑦2

2𝜎2  (5) 

𝜎 = standard deviation distribution (𝜎 = 1,4) 

e = exponential value (e = 2,71828183) 

(x,y) = the midpoint of the Gaussian function 

• Convolution with the Canny method is done by equation 

1

159

(

 
 

2 4 5 4 2
4 9 12 9 4
5 12 15 12 5
4 9 12 9 4
2 4 4 4 2)

 
 

 (6) 

•   Calculate magnitude gradient using equation 

𝐺 = |𝐺𝑋|+ |𝐺𝑦|  (7) 

•   Determining the edges by using the equation 

𝑞 = 𝑡𝑎𝑛−1 (
𝐺𝑥
𝐺𝑦
) (8) 

•   Gives the threshold value of the threshold to the ED by using equations 

𝑔(𝑥,𝑦) = 𝑓(𝑥) = {
1, 𝑖𝑓 𝑇𝑚𝑖𝑛, ≤ 𝑓(𝑥,𝑦) ≤ 𝑇𝑚𝑎𝑥

0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
 (9) 

The results from image processing in canny method can be seen in Fig. 6. 

  

Fig. 6: Edge detection image result. 

249



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

d. Low Pass Filter Using Bilateral Filter 

Low Pass Filter is a method by using a filter type that passes low-frequency 

frequencies and dampens high-value frequencies in imagery [15]. In the process, 

LPF uses the Bilateral filter method. The BF method is a noise reduction filter on 

the ash image and also maintains edge detection for smoother image results [16]. 

Benefits of using the Bilateral filter method include: 

• Maintain edge detection while eliminating image disturbance and refining the 
resulting image 

• Eliminate the opacity of the color image 

• An easy method to apply [16]. 

The Bilateral filter method is calculated using the kernel filter equation 

 𝐾 =
1

𝐾𝑠𝑖𝑧𝑒.𝑤 𝑥 𝐾𝑠𝑖𝑧𝑒.ℎ
(

1 1 … 1
1 1 … 1
… … … …
1 1 … 1

) (10) 

K = kernel bilateral filter 

Ksize = wide of blur in the kernel 

w = width image 

h = high image  

The result of image processing in the Bilateral filter can be seen in Fig. 7. 

 

Fig. 7: Bilateral filter image result. 

e. Image Segmentation 

Image segmentation is the process for recognition of objects in images and also 

feature extraction [13]. By doing image segmentation in the final stage, it will 

show the results of objects and backgrounds on the image. Segmentation is done by 

assigning a threshold value (threshold) that applies to all parts of the input image. 

Segmentation is done using the following equation 

𝑔(𝑥,𝑦) = 𝑓(𝑥) = {
1, 𝑓(𝑥,𝑦) ≥ 𝑇𝑚𝑎𝑥
0, 𝑓(𝑥,𝑦) < 𝑇𝑚𝑎𝑥

 (11) 

 g(x,y) = image after thresholding 

f(x,y) = edge detection image result 

T = limit threshold value 

250



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

 

Fig. 8: Segmentation image result. 

3.2.3 Pin Hole Algorithm Processing 

Determination of the area in the image is the process of determining the area by 

following the length and width of the toll road to be detected. After the image acquisition 

process, area masking is carried out by providing a detection area limit on the toll road 

section. The masking area is done by connecting the four coordinates in the image so that 

it becomes a 2D shape. Then the detection area is calculated using the equation for the 

area of an irregular trapezoid. 

The preprocessed black and white image is divided into a small symmetrical grid (Pin 

Hole). The number of grids used is 8x8, 16x16, 32x32, 64x64 pixels. The process of 

dividing into a grid is done when inserting the CLAHE filter into the grayscale image. 

The next step is to find the centroid point (middle value) in the image in each grid 

area (Pin Hole). To calculate the vehicle capacity in the area to be detected, it can be done 

using Background Elimination (BE) [17]. BE is processed by eliminating everything that 

is considered, not including the object in the image so that the value of the object and the 

background of the image will be separated [17]. BE can be done by first calculating the 

mean of the RGB value of the image with the p-value in the selected background image 

using equation 

𝐵(𝑝) = ∑
𝐼𝐵(𝑘,𝑝)

𝑛𝑘
 (12) 

𝐼𝐵(𝑘,𝑝) = pixel color value for point p in frame k  

Then BE from the process   was calculated with equation 

𝐷(𝑘,𝑝) = {𝐼𝐵(𝑘,𝑝) −𝐵 (𝑃)}  (13) 

{.} = value Euclidean Distance between 𝐼𝐵 and 𝐼𝐵 

After the BE value is obtained, the image of the result of the process is only the 

centroid value of the object of the detected vehicle object. To obtain the value of 

congestion capacity in the area specified, then used equation 

𝐺(𝑥) = 1−
𝐷(𝑘,𝑝)

𝐼𝐵(𝑘,𝑝)
   (14) 

G(x) = congestion value 

After doing the calculation on the input image, we will get the result of the number of 

centroid values contained in the box Pin Hole of the filled capacity of the box. 

251



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

3.2.4 Classification 

Results classification is an advanced process of calculating the number of centroids 

from the Pin Hole area by grouping the calculation results into several categories 

determined based on the amount of capacity of the detected object. According to Rudi 

Herianto (2018), as the Head of the Satlatas Bojongsoang Unit, the congestion category 

can be grouped into 3, namely: 

a.  Traffic Jam 

The jam status is used when the resulting value of the congestion capacity 

calculation reaches a value of more than 70%. 

b.  Crowded Smoothly 

The current status is smoothly used when the value of the result of the congestion 

capacity calculation reaches a value of more than 10% and less than 70%. 

c.  Fluent 

The current status is used when the value of the calculation of congestion 

capacity reaches less than 10%. 

4.   RESULTS 

4.1  Testing the Conversion of Pixel Values into Meter Units 

Testing the conversion of pixel values into units of meters aims to see the effect of the 

results of the detection performed using imagery with the existing conditions significantly. 

To do this test needs to be done in several stages by taking several image samples that 

have a smooth and crowded classification smoothly. The stages that need to be done can 

be described as follows: 

a. Determine the true value of the actual detection 

The actual detection area value is obtained by calculating the area of the road that 

is on the Moh. Toha Toll Road based on Toll Road Geometry Standard [18].  

b. Calculate the actual detection area  

The image ROI is calculated using the formula of building a free trapezoidal space. 

Where both sides of the trapezoid are the length of the toll road and the height of 

the trapezoid is the width of the toll road. 

c. Calculate pixel area and conversion to meters 

Calculating the area of pixels then converting them into meter units is done by 

calculating the area obtained from the image of the detection result into the 

equation 

𝑎𝑟𝑒𝑎 = 𝐿𝑃 ×𝑑𝑝𝑖 ×𝑝𝑖𝑥𝑒𝑙 𝑡𝑜 𝑐𝑚    (15) 

 dpi = dot per inch 

 pixel to cm = 0.0264583333 cm 

d. Determine the Image Scale Value 

Determine the value of the image scale by dividing the original area by the area of 

the pixel area converted into meters. It is helpful as a comparator value between 

the image of the detection result and the condition on the actual toll road. 

252



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

Based on the process of testing the conversion of pixel values to meters, the results are 

obtained for the actual detection area as shown in Table 1. The result of the pixel value 

conversion to the unit meter can be seen in Table 2. After getting each of the required 

values as in Tables 1 and 2, conclusions can be drawn based on the sampling of the 

detected images. Conclusions can be seen in Table 3. 

Table 1: Wide actual toll roads 

No. Explanation Value 

1. DPI 72 pixels per inch camera 

2. Actual Toll Road Length 91 m 

3. Actual Toll Road Width 10.5 m 

4. Actual Toll Road Area 955.5 m2 

 

Table 2: Scale and area of conversion 

No. Area (pixels) Area of 

Conversion 

Area (𝒎𝟐) 

Actual Toll 

Road Area 

(𝒎𝟐) 

Conversion Area 

Scale with Real 

Area 

1. 237900 pixels 4531.93  955.5  4:1 

 

Table 3: The final conversion result to the pixel value 

No Pixel Value Value After Conversion (𝒎𝟐) 

Percentage 

of 

Congestion 

Area Object 

Area 

Percentage 

of 

Congestion 

Classi-

fication 

Area Object 

Area 

1. 237900 17900 7.52% Fluent 4531.93  340.9907 7.52% 

2. 237900 34800 14.62% Smoothly 4531.93  662.939 14.62% 

 

4.2  Pin Hole Grid Testing  

The value of detection accuracy for 20 frames can be seen in Fig. 9. Pin Hole grid test 

aims to see the effect of the number of Pin Hole boxes during the image segmentation 

process. This step is used to detect congestion when segmentation in the image entered in 

the CLAHE filter on the accuracy obtained in the process. CLAHE filter was used because 

the filter input image is divided into the symmetrical grid where the number of grids 

affects the Pin Hole's accuracy. 

a. Image Conformity Results 

This test is done by running 5 different Pin Hole grid values, i.e., grid images 

8x8, 16x16, 32x32, and 64x64 with the image without grid division. The grid 

value used for the original image and the segmentation image of the CLAHE 

filter results. The following results from the process can be seen in Table 4. 

Based on the segmentation image conformity using CLAHE filters with the 

original image in Table 4, the results of segmentation images have a lot of noise 

due to the inadequate CLAHE filters caused by the small number of grids that 

cause excess calculation results when the process is done The Histogram 

Equalizer that causes the calculated vehicle object does not match the input 

253



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

image. Many noises in the results image are obtained in the image with a grid of 

8x8, 16x16 and 32x32. 

 
Fig. 9: Value of accuracy to the number of grids. 

Table 4: Image conformity results 

No. Grid 

Amount 

Results Similar Not 

Similar 
Original Image CLAHE Segmentation 

Image 

1. 0x0 

  

  

2. 8x8 

  

  

3. 16x16 

  

  

4. 32x32 

  

  

5. 64x64 

  

  

55% 55% 55%

75%

85%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

8 x 8 16 x 16 32 x 32 No Grid

Distribution Image

and non CLAHE

Filter

64 x 64

Accuracy Value Against Grid Amount

254



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

a. Image Detection Results 

This was done by comparing the detection value of the image with the correct 

value to get the accuracy value. The correct value is obtained from interviews 

conducted with Mr. Rudi Herianto is the Chief of the Indonesian National Police 

Traffic Department in Bandung.  

4.3  Light Intensity Testing 

Testing will be carried out by taking sample data on the results of recording traffic 

conditions on the toll road with different light intensity values. Different light intensity 

values will be obtained by detecting at different times, i.e., morning, afternoon, evening, 

and night. After that, the comparison of results will be made based on the accuracy values 

obtained. From this stage, the conclusion will be drawn based on high accuracy values on 

the value of the light intensity obtained by concluding the vulnerable value of the 

appropriate light intensity to detect congestion. The results of this test can be seen in Fig. 

10. 

 
Fig. 10: Value accuracy of results based on light intensity. 

Based on the graph in Fig. 10,  the value of light intensity affects the result of the accuracy 

value obtained from the detection of congestion. The difference in the value of accuracy 

obtained is based on several factors as follows: 

a. Toll Road Conditions 

As seen in Fig. 10 that the detection error that occurred on the Toll road segment 

at light intensity is 11500 with an accuracy value of 0%. This is due to the 

detection of the image background as an object; after an analysis of the original 

image and the segmentation results obtained results that the error lies in the color 

of the uneven markers. 

b.  Car Light 

Although in Fig. 10 the detection at night, the accuracy value reaches 100%, 

because the results of true values obtained from subjective opinions do not 

guarantee that the results are always correct. Reflection of light from car lights is 

considered to have an excessive amount of light intensity from the specified 

threshold. Therefore, the value of the excess light intensity is detected as an 

0%

60%

90%
100%

0%

20%

40%

60%

80%

100%

120%

11500 41379 5430 93

Value of Results Accuracy

255



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

object when image segmentation and is calculated in the search for congestion 

capacity values. 

5.   CONCLUSION 

Based on the results of the analysis of the testing scenario on the congestion detection 

system using image processing with Pin Hole algorithm, which is carried out on 5 toll 

roads that will enter the city of Bandung, the following conclusions are obtained: 

1.   Based on the conversion of pixel values into units of meter obtained results from 
the image of the sample taken that the area of detection value using the image of 

237900 pixels equal to the area of 4531.93 m2 on the actual condition with a scale 

of 1:4 

2.  The congestion capacity for the classification results smoothly in the sample image 

will be the same value when the total object value of 340.9907 m is in the area of 

the detection area of 4531.93 m2 in the real condition. 

3.  The implementation of the system can produce an accuracy value of 85% by using 

symmetrical 64x64 grid images with an average computing time of 5 seconds. 

4.  The detection results are assessed in the light intensity range from 5.430 to 41.379 

LUX, assuming good road conditions. 

5.  The algorithm used still require modification with other algorithms in order to do 

detection at night. 

6.  Based on the weather obtained when sampling is not so influential as long as the 

light intensity values obtained are still in the susceptible 5.430 to 41.379. 

7.  This system is helpful as an alternative to find the value of congestion capacity 

based on the number of vehicles that will enter the city of Bandung through the 

highway. 

REFERENCES  

[1] Royani, Tahere, Haddadnia J, Pooshideh MR. (2010) A simple method for calculating vehicle 
density in traffic images. In 6th IEEE Iranian Conference on Machine Vision and Image 

Processing: October 2010; pp 1-4.   

[2] Fibriliyanti, Yanita, Faradila LR, Taqwa A. (2017) Implementasi Pengolahan Citra dengan 
Metode Histogram Of Oriented Gradient (Hog) untuk Pengaturan Waktu pada Traffic Light 

Berdasarkan Deteksi Kepadatan Kendaraan. In Prosiding SNATIF: 2017; pp 403-412. 

[3] Directorate General of Land Transportation Ministry of Transportation Republic of Indonesia. 
(2008) Technical Instructions for the Implementation of Road Equipment. Jakarta. 

[4] Nurhadiyatna A, Hardjono B, Wibisono A, Sina I, Jatmiko W, Ma'Sum MA, Mursanto P. 
(2013) Improved vehicle speed estimation using gaussian mixture model and hole filling 

algorithm. In 2013 IEEE International Conference on Advanced Computer Science and 

Information Systems (ICACSIS): September 2013; pp 451-456. 

[5] Kim KI, Jung K, Kim JH. (2002) Color texture-based object detection: an application to 
license plate localization. In Springer International Workshop on Support Vector Machines: 

August 2022; Berlin, Heidelberg. pp 293-309. 

[6] Bas E, Tekalp AM, Salman FS. (2007) Automatic vehicle counting from video for traffic flow 
analysis. In 2007 IEEE intelligent vehicles symposium: June 2007; pp 392-397. 

[7] Mahatmaputra S, Permata E, William. (2011) Deteksi kemacetan lalu lintas melalui kamera 
menggunakan pin hole algorithm. ComTech., 2:821-834. 

256



IIUM Engineering Journal, Vol. 23, No. 1, 2022 Irawanto et al. 
https://doi.org/10.31436/iiumej.v23i1.2135 

 

[8] Cucchiara R, Massimo P, Paola M. (2000) Image analysis and rule-based reasoning for a 
traffic monitoring system. IEEE transactions on intelligent transportation systems, 1(2):119-

130. 

[9] Chen TH, Lin YF, Chen TY. (2007) Intelligent vehicle counting method based on blob 
analysis in traffic surveillance. In Second IEEE International Conference on Innovative 

Computing, Informatio and Control (ICICIC): September 2007; pp 238-238. 

[10] Zhang H, Liptrott M, Bessis N, Cheng J. (2019) Real-time traffic analysis using deep learning 
techniques and uav based video. In 16th IEEE International Conference on Advanced Video 

and Signal Based Surveillance (AVSS): September 2019; pp. 1-5. 

[11] Saepudin D, Rizal A. (2014) Analisis Contrast Limited Adaptive Histogram Equalization 
(Clahe) Dan Region Growing Dalam Deteksi Gejala Kanker Payudara Pada Citra 

Mammogram. Jurnal Elektro Unika Atma Jaya., 7(1):15-28. 

[12] Reza AM. (2004) Realization of the contrast limited adaptive histogram equalization 
(CLAHE) for real-time image enhancement. Journal of VLSI signal processing systems for 

signal, image and video technology., 38(1):35-44. 

[13] Junior RA, Nurhasanah N, Sanubary I. (2014) Perbandingan Penggunaan Beberapa Metode 
Deteksi Tepi Pada Pengolahan Citra Radiologi Fraktur Tulang. Prisma Fisika., 5(3):122-130. 

[14] Triono P, Murinto M. (2015) Aplikasi Pengolahan Citra Untuk Mendeteksi Fraktur Tulang 
Dengan Metode Deteksi Tepi Canny. Jurnal Informatika Ahmad Dahlan., 9(2):103-842. 

[15] Panda DK, Sukadev M. (2013) A gaussian mixture model with gaussian weight learning rate 
and foreground detection using neighbourhood correlation. In 2013 IEEE Asia Pacific 

Conference on Postgraduate Research in Microelectronics and Electronics (PrimeAsia): 

December 2013; pp. 158-163. 

[16] Tomasi C, Manduchi R. (1998) Bilateral filtering for gray and color images. In Sixth IEEE 
international conference on computer vision (IEEE Cat. No. 98CH36271): January 1998; pp. 

839-846. 

[17] Deswal S, Gupta S, Bhushan B. (2015) A survey of various bilateral filtering techniques. 
International Journal of Signal Processing, Image Processing and Pattern Recognition., 

8(3):105-120. 

[18] Atkočiūnas E, Blake R, Juozapavičius A, Kazimianec M. (2005) Image processing in road 
traffic analysis. Nonlinear Analysis: Modelling and Control., 10(4):315-332. 

[19] The Directorate General of Directorate General of Marga The Ministry of Public Works and 
Housing Republic of Indonesia. (2008) Standar geometri jalan bebas hambatan untuk jalan 

tol. Jakarta. 

 

 

257