Research On Image Repair And Recognition Method Of Damaged License Plate Based On Machine Vision

To solve the adverse impact of a large number of cars on traffic,China began to apply intelligent transportation system,license plate recognition is an indispensable part of the system.At present,license plate recognition technology is at a relatively mature stage,but the general license plate recognition system can’t accurately recognize the license plate with dirt stains and character wear.In order to solve this problem,the image repair and character recognition of the damaged license plate image can be studied,so as to improve the recognition correct rate of damaged license plate.The damaged license plate was taken as the research object,using the background color and the external frame’s characteristics of the license plate,and the morphological processing and edge detection related algorithms were combined to locate the complete license plate image.In view of the situation that there was an error in the extraction of the damaged part of image,using image segmentation algorithm,a Fuzzy C-Means algorithm was designed that can automatically determine the number of clusters,and it was used to segment the color of the image to solve this problem.Because Criminisi algorithm couldn’t achieve the ideal state of the repair effect of license plate image,the priority calculation,the size of the matching blocks and the search method of matching blocks in the algorithm were improved,it was used to restore the damaged parts,so that the characters of the license plate image to restore integrity.In the recognition method of template matching,similar characters were often confused,a method of classifying template libraries and comparing character subgraph with template subgraph was proposed to find the most suitable matching template and complete character recognition.The experimental results show that this method improves the repair effect and speeds up the repair speed.The NIQE value of the image has decreased and the repair time has been reduced to about 1/2 of the original time.At the same time,the recognition rate of damage license plate characters has been improved to 90%.Figure 36;Table 4;Reference 65...