Research on License Plate Recognition System Based on MATLAB

1 Introduction

The vehicle license plate is the only management identification symbol of a motor vehicle and plays an irreplaceable role in traffic management. Therefore, the vehicle license plate recognition system should have a high recognition rate, a large tolerance threshold for factors such as ambient lighting conditions, shooting position and vehicle speed, and should meet real-time requirements. Automatic license plate recognition is a pattern recognition technology that uses dynamic video or static images of vehicles to automatically identify license plate numbers and license plate colors. Its hardware generally includes triggering, camera, lighting, image acquisition and other equipment, and its software core includes algorithms such as license plate positioning, license plate character segmentation and character recognition.

2 System Implementation

2.1 System Overview

A complete license plate recognition system should include vehicle detection, image acquisition, image preprocessing, license plate location, character segmentation, character recognition and other units. When the vehicle reaches the trigger image acquisition unit, the system acquires the current video image, the license plate recognition unit processes the image, locates the license plate position, and then segments the characters in the license plate for recognition, and then forms the license plate number output. The principle of the license plate recognition system is shown in Figure 1.

2.2 Image Preprocessing

The input color image contains a lot of color information, which will take up a lot of storage space and reduce the execution speed of the system when processed. Therefore, when performing image recognition and other processing, the color image is often converted into a grayscale image to speed up the processing speed. The main MATLAB statements for grayscale processing, background image extraction, enhancement processing, image binarization, edge detection, filtering, etc. are as follows:

2.3 License Plate Location

In the natural environment, the background of the car image is complex and the lighting is uneven. Accurately determining the license plate area in the natural background is the key to the entire image recognition process. First, a large-scale correlation search is performed on the collected image to find several areas that meet the characteristics of the car license plate as candidate areas, and then these candidate areas are further analyzed and judged. Finally, an optimal area is selected as the license plate area and it is segmented from the image. At the same time, the problem of license plate tilt must be considered. The algorithm flow is as follows:

(1) Perform region extraction on the binary image, calculate and compare the regional feature parameters, and extract the license plate area.

(2) Calculate the minimum width and height of the marked area, and based on previous knowledge, extract and display a closer license plate binary value map.

(3) Solve the problem of license plate tilt by calculating the license plate rotation angle. The license plate tilt causes the peaks and valleys of the projection effect to be unclear, so the license plate needs to be corrected. The linear fitting method is used to calculate the angle between the fitting line at the point where the image value of the upper or lower side of the license plate is 1 and the horizontal X-axis. The license plate rotation angle and the license plate binary sub-image after rotation and binarization are calculated using the MATLAB function Imrotate. The processing results are shown in Figure 2.

2.4 Character Segmentation

After the license plate area is located, the license plate area is segmented into individual characters. The vertical projection method is generally used. Since the projection of the character in the vertical direction must obtain a local minimum at the gap between characters or within characters, and this position should meet the conditions of the license plate character writing format, character, size restrictions, etc. The vertical projection method is used to achieve better character segmentation in car images under complex environments. By analyzing and calculating the horizontal and vertical projections of the characters, the license plate character height, the top and bottom lines of the characters, the character width, and the center position of each character can be obtained to facilitate the extraction of segmented characters. Then calculate the vertical projection of the license plate and remove the vertical border of the license plate. Get the average width of the license plate and characters. Finally, calculate the center position and maximum character width of each character of the license plate, and extract the segmented characters. The algorithm flow is shown in Figure 3. The license plate character height and width calculated by the program algorithm and the segmented characters are shown in Figure 4.

3 Conclusion

Judging from the results of MATLAB programming, the image recognition algorithm used here is very effective in locating license plates. The algorithm can effectively detect the upper, lower, left, and right borders and rotation angles of license plate images, and accurately segment and recognize license plate characters. Through experiments on multiple license plates, the accuracy rate is high. Compared with the traditional C++ language, the workload and development cycle are greatly reduced. In practical applications, the recognition rate of the license plate recognition system is closely related to the quality of the license plate and the quality of the image shooting. It is also affected by various factors, and the recognition system and algorithm need to be continuously improved.