Research And Realization Of License Plate Location Algorithm

With the rapid development of the country's economy, the improvement of people's living standard and the increasing popularity of the highway, automation of the urban traffic management has become a problem which should be dealt with urgently. Intelligent Transportation System (ITS) is not only an effective solution to traffic congestion problems, but also plays significant role in emergency treatment of the city traffic, environment protection and energy saving. License plate recognition system (LPRS) is the most important part in ITS, with license plate location part as the key of the entire identification system.Two location algorithms are studied in the present thesis. One is based on the color information of the license plate and the other is based on mathematical morphology methods. The plate location algorithm based on the color information of the license plate uses a color-binarization algorithm first. Then it searches the coordinates of the four corners in the license plate region. The plate location algorithm based on mathematical morphology extracts the peaks in the images with Top-hat method, binaries them with OTSU algorithm, and extracts the edges with an improved Sobel algorithm. Finally it generates connected regions of the images, searches the license plate region and locates the region of the license plate. The advantages and disadvantages of the two location algorithms are analyzed and compared through experiments. The two algorithms are used in combination according to the application needs, which can achieve the highest positioning accuracy and least time complexity.With the precise license plate location, Hough transform and Radon transformation are analyzed and the result shows that the latter is more suitable for the tilt detection in this paper. Then steps for locating the upper, lower, left and right boundary of the license plate are introduced, which includes locating the upper and lower boundary with a white-black-jumping method, locating the left and right boundary with the color information and texture of the license plate. They are verified by experiments.Based on the study of license plate location on static images, license plate location methods on video sequences are further explored. Improvements are made on the algorithm based on frame difference. Statistics of the experimental results is given.Finally, a license plate location software system is designed and developed. The proposed algorithms in this paper are compared and tested, the result of which shows that the correct location rate and the time consumption of the system have reached the application standard.