Review Paper: Licence Plate and Car Model Recognition

  • R. Akshayan
  • S. L. Vishnu Prashad
  • S. Soundarya
  • P. MalarvezhiEmail author
  • R. Dayana
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 768)


The main intention of this study paper is to explore and analyze the numerous methods that are used for license plate extraction and identification. Analysis is done by detailed study of the prevailing methodologies and their drawbacks were duly notified. Various concepts such as OCR reading, bounding method and other computational techniques were thoroughly studied and are recorded below. Thus, with the knowledge gained from this review, a pristine and a noble method were developed for the same by utilizing the concept of API. This not only involves extraction of characters from the license plate but also aids in the identification of the vehicle model, as a whole, with the help of image processing. This method is not merely effective for detection but also holds good for certain amount of automation with the aid of the upheld database. These techniques of detection backboned by databases can prove to be very effective in a wide range of use-cases in the current world. Such identification as well as compilation of useful data can be implemented in applications such as automatic toll system, fee collection in parking areas, traffic disciplinary maintenance, vehicle monitoring inside a premise and much more.


License plate Vehicle identification Automatic toll system Automated parking fee collection API 

1 Introduction

License plate extraction and model identification [1, 2, 3, 4] of a vehicle has immense importance in various issues. Implementation of such system is mainly done in automatic toll collection [1], parking fee collection, etc. The earliest version of automobile identification was done with microwave and infrared system [1]. Over the course of time, this process was further extended and using a CCD camera, image segmentation of license plate was processed [2]. With advent of capturing technology, a method called Optical character recognition was devised to extract plate information under various circumstances [5]. This includes places where an image of a vehicle in traffic has to be processed or in cases, where a vehicle has to be identified within a mixture of other objects. Texture analysis is applied along with statistical methods [6] to provide information that is obtained from traffic camera images. Usage of Tesseract engine and neural networks [7] has given more efficient character segmentation and identification. Based on prior knowledge on the images, an automated detection which detects only the edges of license plate was performed [8]. This process helps to identify merely the license plate with the help of above-mentioned process. Reduction of noise can be avoided by the process of converting a colored image to a gray scale image. To allow recognition of each character, segmentation can be done by yet another method [9]. In the same method, an image was converted to YDbDr format to specifically detect characters [10]. The automated fee calculation was thus done using these recognition techniques, along with certain prescribed rules that was controlled via a program [11].

1.1 Various Techniques Involving Extraction of Characters from License Plate

Extraction of license plate characters has been extensively researched for various purposes. Some of the effective theories have been put forth in this section.

1.1.1 Optical Character Recognition [OCR]

As put forth in papers mentioned [7, 8, 11], the OCR technology is used for segmentation of the characters that are available in the license plate. The process involves capturing and extraction of the image pertaining only to the license plate, as shown in Fig. 1. The image so obtained is segmented by character, which is then recognized individually, using the OCR process.
Fig. 1

System using OCR recognition

The process of such image capturing technology is different with those mentioned in [7]. Here, all the operations are programmed in a mobile device. This mobile device, which either runs on android or iOS, is used to capture the image with its camera. Then on, with the help of OpenCV libraries and image processing techniques, further process in the system can be performed. In paper [8], for better recognition purpose, the process of machine learning is utilized. By using supervised machine learning, various types of plates are studied whose data are stored in XML. This process could facilitate the system from differentiating the license plate from other objects. Nevertheless, the main disadvantage of this method lies in its usage of XML. Since, they are large data storage units, they take up high disk space, which in turn overburden the system. Moreover, analyzing paper [11], this indicates a similar process except for the use of gray scaling, given by the Eq. (1) as,
$$ Y = 0.299R + 0.587G + 0.114B $$
Whereas in case of [8], binary form of image is obtained, as shown in Fig. 2. This results in possibility of error due to improper detection of details.
Fig. 2

License plate localization

1.1.2 Using Bounding Method

Analyzing this method [5], implies that, even though OCR is applied for preliminary process, the recognition block, which identifies the edge of the license plate, varies. The image obtained in this case is extracted from the overall vehicular image. Thus, instead of processing whole image as in [9], only the license plate is isolated for further processing. This method is illustrated in Fig. 3. As a result, the computational error can be minimized, which can be seen from the Table 1. Despite these advantages, this method cannot be applied for complex image detection due to the presence of multiple edges. Thus, additional complexity is introduced due to this process.
Fig. 3

Using bounding method

Table 1

Results of the system [9]


Number of accuracy

Percentage of accuracy (%)










1.1.3 Extraction Process for Characters

The system suggested in paper [2] provides the process of masking the unwanted components obtained from the image through camera.

This system, as shown in Fig. 4, consists of a simple unit wherein the arrival of the car is identified with the help of IR sensor, upon which the gate would be closed. Thus, the image of the plate can be captured and processed. The quality of the image is maintained by fixing a threshold value to eliminate noise components. The threshold value may vary with different parameter such as,
Fig. 4

Process of extraction using sensor

WA: width-threshold; HA: height-threshold; AA: area-thresholdmax

Modified Adaptive thresholding is utilized for segmentation process. This method proves its effectiveness in terms of lesser processing time and memory. However, it produces lesser computational power when compared to OCR.

1.1.4 Computation for Multiple License Plates

All the above suggested techniques are effective and adaptable for a clear image of single license plate. But in a real-time scenario, a traffic camera can obtain images with certain resolution, only for a certain number of images. In such cases, the process, become tedious and this may raise the complexity of the processes, thereby making the system prone to error. This paper [10] provides a novel method for extraction of characters from numerous license plates at a single time. From the Fig. 5, it is shown that the process involves converting a RGB image to YDbDr format. This system was designed specifically for China where the number plates are of either blue or yellow (for heavy vehicles). Thus, with such high precision, it can be implemented in various important places. Nevertheless, the major shortcoming in this method is the requirement of high quality image thereby increasing the cost of the system. Likewise, in yet another paper [6], similarly, the image from the traffic cam is identified by probabilistic feature detection. But the main drawback in this method is to achieve the correct probability and to overcome the increased complexity when dealing with multiple plates.
Fig. 5

Detection of multi-colored license plate

2 Suggested Method for License Plate Character Extraction

A suggested method, as shown in Fig. 6, involves integration of API with a motion detector sensor. With advancement of technology in recent years, Google engine have let out various improvised version of resources for innumerable purposes. Utilizing these resources has made computational analysis with stored database much simpler. With the aid of motion sensor, the arrival of vehicle can be detected and with optimum resolution, an image is taken from the front and rear end of the car. These images are further processed with the help of Google API, which in turn provides the required result with more accuracy. This is done by integrating the camera with Raspberry pi, which processes and then sends this image to Google-Vision-API. By suitable programming, the text can be retrieved from the image upon which the characters are extracted. Using a SQL database, the data is stored. When required, the data can be extracted, deleted or inserted as and when required. Unlike XML, the data is not redundantly large and thus computation space is further conserved.
Fig. 6

Proposed method of detection and extraction

3 Summary

Thus, various methods were analyzed pertaining to license plate detection and recognition. Each method was studied thoroughly and were estimated and compared on their performance basis. By considering all the shortcomings in the subsisting systems, a novel method of image recognition was put forth for license plate detection along with car model recognition. The proposed method was tested in terms of performance, operating speed, cost and its effectiveness to be implemented in real-life situations. Under all these criteria, the proposed method of detection involving API provides precise output with a sturdy database to support in various applications. With the help of faster processing system by raspberry pi, computational error can also be minimized. Thus, by means of simpler programming structure, it can be made adaptable to various use-cases.


  1. 1.
    Foote, R.S.: Automatic vehicle identification: tests and applications in the late 1970’s. IEEE Trans. Veh. Technol. 29(2), 226–229 (1980)Google Scholar
  2. 2.
    Wu, B.-F., Lin, S.-P., Chiu, C.-C.: Extracting characters from real vehicle license plates out-of-doors. IET Comput. Vis. 1(1), 2–10 (2007)Google Scholar
  3. 3.
    Liu, G., Ma, Z., Du, Z., Wen, C.: The calculation method of road travel time based on license plate recognition technology. In: Proc. Adv. Inform. Tech. Educ. Commun Comput. Inform. Sci., vol. 201, pp. 385–389 (2011)Google Scholar
  4. 4.
    Kranthi, S., Pranathi, K., Srisaila, A.: Automatic number plate recognition. Int. J. Adv. Tech. 2(3), 408–422 (2011)Google Scholar
  5. 5.
    Du, S., Ibrahim, M., Shehata, M., Badwy, W.: Automatic License Plate Recognition (ALPR): a state-of-the-art review. IEEE Trans. Circ. Syst. Video Technol. 23(2), 311–325 (2013)Google Scholar
  6. 6.
    He, H., Shao, Z., Tan, J.: Recognition of car makes and models from a single traffic-camera image. IEEE Trans. Intell. Transp. Syst. 16(6), 3182–3192 (2015)Google Scholar
  7. 7.
    Do, H.N., Vo, M.T., Vuong, B.Q., Pham, H.T., Nguyen, A.H., Luong, H.Q.: Automatic license plate recognition using mobile device. In: 2016 International Conference on Advanced Technologies for Communications (ATC)Google Scholar
  8. 8.
    Miyata, S., Oka, K.: Automated license plate detection using a support vector machine. In: 2016 14th International Conference on Control, Automation, Robotics & Vision Phuket, Thailand, 13–15th Nov 2016 (ICARCV 2016)Google Scholar
  9. 9.
    Mahesh Babu, K., Raghunadh, M.V.: Vehicle number plate detection and recognition using bounding box method. In: 2016 International Conference on Advanced Communication Control and Computing Technologies (ICACCCT)Google Scholar
  10. 10.
    Asif, M.R., Chun, Q., Hussain, S., Fareed, M.S.: Multiple licence plate detection for Chinese vehicles in dense traffic scenarios. IET Intell. Transp. Syst. 10(8), 535–544 (2016)Google Scholar
  11. 11.
    Yimyam, W., Ketcham, M.: The automated parking fee calculation using license plate recognition system. In: International Conference on Digital Arts, Media and Technology (ICDAMT) (2017)Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2019

Authors and Affiliations

  • R. Akshayan
    • 1
  • S. L. Vishnu Prashad
    • 1
  • S. Soundarya
    • 1
  • P. Malarvezhi
    • 1
    Email author
  • R. Dayana
    • 1
  1. 1.SRM Institute of Science and TechnologyKattankulathur, ChennaiIndia

Personalised recommendations