You are on page 1of 4

BENC 4483 : Design a Computer Vision System

for a Recognition Application


Nur Ariana binti Zainurin
Fakulti Kejuruteraan Elektronik dan Kejuruteraan Komputer (FKEKK), Universiti Teknikal Malaysia Melaka (UTeM),
Hang Tuah Jaya, 76100, Durian Tunggal, Melaka, Malaysia.

B021710185@student.utem.edu.my

Abstract — The License Plate Recognition (LPR) is computing compared to most conventional methods. Within
nowadays one of the most popular technologies in the Smart a fraction of a second (less than 0.5 seconds), the proposed
Transport System (ITS). The software systems will recognise system can recognize any type of number plate.
the license plate number on their own using this technology Abhishek Kashyap, et. al in his proposed studies of
from a digital image. This assignment discusses the method
series for recognizing the license plate number. The methods
automatic number plate recognition[4], such an ANPR
are a grayscale converting the colorful picture into gray, a gadget is unavailable outside the shelf, it is their goal to
binary image which transforms the grayscale image further personalize an ANPR system for instructional institutions.
into a black and white version, flat number detection to check Templates can be used on static photos and an average
for the platform number separation character segment which accuracy of 82.7 per cent is obtained. The execution of the
separates the extracted characters and pixels into meaningful given method can be further pushed by the use of multi-
information. level genetic algorithms for the popularity of the number
plates of multiple vehicles in a single photo body. Also, an
I. INTRODUCTION AND REVIEW OF ARTICLE extra easier variant of this gadget can be done by capturing
photos from the stationery clip and selecting a large car
In the theme of image processing applications, a border for the category of vehicles and spotting the quantity
license plate recognition system is common. The main plates.
reason is based on the application side of the system, such as A successful and rapid detection process for
e-parking space management systems, vehicle traffic multiple Indian and Spain license plates. For location
violation detection systems, as well as the stolen vehicle detection of the number plate, Automatic License Platform
identification systems. Recognition (ALPR). High accuracy in the plate detection
In contemporary license plate recognition (LPR) part is our advantage on several plates. In a single-camera
systems, there are many recognition approaches used. Most frame, the method proposed can detect multiple number
of them consist of three main measures: localization of plate vehicles. The license with a complex background is
license plates, segmentation of characters, and identification correctly monitored and a good result has been achieved[5].
of optical characters (OCR). Each of these steps must have B. Pechiamma, in the proposed strategy, is linked
high rates of success to increase the reliability and accuracy and tested in various images in this project and contrast to
of the entire system[1]. deferentially with the time and precision of recognition. In
Thus, as long as you have a fixed monitor set up this proposed work, a broad study has been carried out to
with the license plate recognition system, which is directed establish an exact recognition framework for numeric plates.
at a specific angle, you can get the approximate location of In contrast to past strategies, OCR produces less time and
the license plate image in advance and then return and then remarkable accuracy for template compatibility recognition.
calculate the obtained results. This shows that recognition speed is greatly improved[6].
Chuin-Mu Wang in his proposed system of plate The automatic vehicle identification system is
number recognition system using iterative binarization presented in this paper[7]. The system uses image
method[2]. This article proposes systems for recognition of processing techniques to detect the vehicle from the PC-
license plates. On the pre-processing phase, set the area of saved database. The system is installed in Matlab and
interest, so that the system icon size will simplify many in performance in real images is tested. The simulation results
future processes. With the Convergence License Plate show that the system robustly identifies and identifies the
Detection Phase, iterative design a set of condition binaries vehicle with the use of a license plate against various
processes to help to change the result of image binarization. lighting conditions. However, there is still room for
The location of the license plates, the licensed device improvement in the implementation. Due to the long shutter
characteristics minimize speeds and improved output time, the camera used in this project is sensitive to
performance. The main algorithm for putting the license vibrations and rapidly change targets. When using a high-
plate number, on the other side. Concepts of the neural resolution camera, the system robustness and speed can be
network convolution, as an adaptive classifier. increased. The OCR methods used for the recognition in
In [3], this paper provides an efficient first this project are sensitive to distortions and different sizes,
calculation technology to identify the number plate of the and affine transformation can be used to improve the
vehicle. In this technique, the algorithm uses a limited identification of OCR from various dimensions and angles.
number of calculations. It is therefore very cost-effective in The statistical analysis can also be used to determine the

1
probability of the number plate detection and recognition. pixels and containing some black pixels in the central area
[8][1][9][10] to detect the Region of Interest (ROI).

Character segmentation
I. THE PROPOSED SYSTEM
These algorithms are normally done by reducing image
information's complexity, thereby improving the functioning
A. Flowchart of the system of the next step system by accelerating the processing speed
of the next steps or by restricting the detection range and at
the same time maintaining the sample destination. In
There is a total of seven stages to process the plate addition, the system allows the segment characters to be
number recognition. The process is shown in Figure 1. separated from the cropped plate number. As the number of
the plate white is white and black in background colors, only
the white pixels are found in the image.

Character recognition

Character recognition is the last feature of this solution. The


Start program uses a similar framework for character recognition.
The 0-9 and alphabet A-Z character template is stored in an
image folder on the system. Both the segment character and
Image binarization the character template are converted into an array by the
system. The pixel is compared to the color by means of a
display. As the picture only contains black and white pixels,
Plate number detection the color of the pixels in both pixels is not identical.

II. RESULTS
Character segmentation
A total of 6 images were used and tested from the beginning
of gray scale up to the end of character recognition to test
Character recognition the performance of the entire recognition process. Each
image is in different length and environment. For number
plate recognition phase are used only for Malaysia plate
Output display number only.

End

Figure 1: Process flowchart

Image binarization

Due to the fact that the calculation of license plate location


in the system results in the mistaken results from some
effects of external factors such as light plate, dirt and
damaged images obtained before treatment in an anticipated
way. The binarization image can be done as the first step in Figure 2: The designed system in MATLAB
the entire license plate recognition process. This feature
enables the system to take the image and turn it into a binary
format with only white and black pixels. The reason for the
Table 1: Success of Number plate recognition
binarization is to facilitate the recognition and
differentiation of the background and the front of the image. No Plate number Plate number recognition
Not only that, since the color of the license plate (black) is . in MATLAB
highly different from the color of a character (white), it will 1.
also be separated and redefined by the binarization process.
However, because of performance concerns about an
unnecessary or unwanted region, only the possible plate
region will be binarised.

Plate number detection

By searching the white area from the binary image, the


system can detect the license plate. The system will attempt
to detect a rectangle with four ranges connected to white

2
2. of an indoor parking lot. As such, it is not suitable for
outdoor use, as the captured picture can be different.
Characters can therefore be obtained in the default format,
i.e. the character gap, and also present several challenges to
the Informal license plate, for example by arranging the
characters too closely between themselves.

REFERENCES

3.
[1] Ping Dong, Jie-hui Yang, and Jun-jun Dong, “The
Application and Development Perspective of
Number Plate Automatic Recognition Technique,”
no. 1, pp. 744–747, 2006.

[2] C. M. Wang and J. H. Liu, “License plate


recognition system,” 2015 12th Int. Conf. Fuzzy
4.
Syst. Knowl. Discov. FSKD 2015, pp. 1708–1710,
2016.

[3] R. Islam, K. F. Sharif, and S. Biswas, “Automatic


vehicle number plate recognition using structured
elements,” Proc. - 2015 IEEE Conf. Syst. Process
Control. ICSPC 2015, no. December, pp. 44–48,
5. 2016.

[4] A. Kashyap, B. Suresh, A. Patil, S. Sharma, and A.


Jaiswal, “Automatic Number Plate Recognition,”
Proc. - IEEE 2018 Int. Conf. Adv. Comput.
Commun. Control Networking, ICACCCN 2018, pp.
838–843, 2018.
6.
[5] A. Menon and B. Omman, “Detection and
Recognition of Multiple License Plate from Still
Images,” 2018 Int. Conf. Circuits Syst. Digit.
Enterp. Technol. ICCSDET 2018, pp. 1–5, 2018.

[6] B. Pechiammal and J. A. Renjith, “An efficient


approach for automatic license plate recognition
This method successfully detected and recognized the plate system,” ICONSTEM 2017 - Proc. 3rd IEEE Int.
number from Malaysia. The method was able to detect and Conf. Sci. Technol. Eng. Manag., vol. 2018-January,
recognize the character from 0-9 and A-Z. The system is pp. 121–129, 2017.
fast, cost-effective and efficient. The final output is shown
in Table 1 above. [7] M. T. Qadri and M. Asif, “Automatic number plate
recognition system for vehicle identification using
III. CONCLUSION optical character recognition,” 2009 Int. Conf. Educ.
Technol. Comput. ICETC 2009, pp. 335–338, 2009.
To be conclude, algorithms are available for the LPR [8] R. M. Sferle and E. V. Moisi, “Automatic number
technology, ranging from photo acquisition to character plate recognition for a smart service auto,” 2019
segmentation to the final optical character recognition. 15th Int. Conf. Eng. Mod. Electr. Syst. EMES 2019,
Every process has its own rationale to achieve the desired pp. 57–60, 2019.
outcome. They are closely linked because the outcome is a
linear process. In other words, they are fully mutually [9] B. Y. Amirgaliyev, C. A. Kenshimov, K. K. Kuatov,
dependent. Binarisation and license plate identification are M. Z. Kairanbay, Z. Y. Baibatyr, and A. K.
the most prominent elements. If the system cannot properly Jantassov, “License plate verification method for
detect the license plate, the rest of the system can become automatic license plate recognition systems,” Proc.
useless. The architecture of this approach simplifies and 2015 12th Int. Conf. Electron. Comput. Comput.
enhances performance compared to other similar systems. ICECCO 2015, pp. 3–5, 2016.
Without an unnecessary processing cycle like noise
reduction, gaussian filters or edge detection, the process is [10] A. Beibut, K. Magzhan, and K. Chingiz, “Effective
simplified since the anticipated result can be achieved algorithms and methods for automatic number plate
efficiently.One of the system's limitations is its flexibility. recognition,” 8th IEEE Int. Conf. Appl. Inf.
The algorithm is developed by taking into account the size Commun. Technol. AICT 2014 - Conf. Proc., 2014.

3
4

You might also like