Research Of Vehicle Detection And License Plate Recognition Algorithm

With the flying economic development and the increasing number of automobiles, the problems of transportation are more serious, so it causes the research on intelligent transportation system (ITS).The technology of vehicle detection and license plate recognition are important parts of intelligent transportation system and hot research focus in image processing field.In intelligent transportation system, vehicle detection can be achieved in information acquisition and process on the moving automobiles, furthermore, license plate recognition can be used in automatic fare collection,contrary vehicles monitoring or parking lot management and so on, it enhances the traffic management efficiency and service capabilities.Firstly, this paper introduces the development situations of intelligent transportation system,some basic theories of image processing and several kinds of classical movement object detection algorithm,then,it elaborates one kind of detection algorithm which is improved and auto-adapted background algorithm. From background model establishment and initial background acquisition, it proposes a new kind of auto-adapted background algorithm. It establishes a mixed Gaussian background model firstly and then it extracts background image by Mid-value clustering. Finally, it updates the background image by using block thought in order to realize the real-time detection of the vehicle.Secondly, it extracts a key frame from the video image sequence and makes a license plate location on the frame .Before location, it should make a pretreatment on this image including image enhancement,binarization and edge detection. It proposes a license plate location methods by using morphology and Projection method. Firstly, it uses opening and closing mathematical morphological processing on the image. Secondly, it extracts license plate area for the first time. Due to the influence of factors including monitoring equipment or road conditions and vehicle movement,the extracting license plate will appear tilt inevitably. Compared with Hough method, it calculates the tilt angle by straight line fitting. Finally, it obtains the license plate for the second time by rotating the license plate.Finally, after license plate location, the next step is eliminating the region of license plate, then carries on the vertical projection to the car license, analyzes related information between the peak value and the valley value in the image, and according to the car license characteristic information, discover the central place of the character, extract and segment the license plate character at last. After license character division, it proposes an improved template matching method to recognize the character. Firstly it establishes template database of letters and characters, then carries on the normalization to the division character. Finally, matched with the template database, it outputs the recognition result.This paper proposes some methods about vehicle detection and license plate recognition, and then carries on the experiment simulation by Matlab. The algorithm has good examination and recognition effect and it has certain theoretical significance and practical value.