Research And Implementation On Algorithms Of License Plate Location And Character Segmentation

With the acceleration of urbanization, the further development of the national economy and the increase of the amount of vehicles, the traffic situation in urban areas is deteriorating. So it is particularly important to establish an intelligent transportation system with information and digital technology. The license plate recognition system is the core element of the intelligent transportation system. The license plate recognition system can be widely used in areas such as toll payment, electronic police, detecting traffic flow and controlling and managing parking and can help improve working efficiency and reduce manpower, which makes it a both economical and practical system. Therefore, the license plate recognition system has greate economic value and practical significance. The license plate recognition system mainly consists of five modules:image acquisition, image preprocessing, license plate location, character segmentation and character recognition. On the basis of predecessors'studies, image preprocessing, license plate location and license plate character segmention are deeply discussed and studied in this thesis.For the vehicle image preprocessing, firstly, the color vehicle image is converted into gray vehicle image by weighted average method and the gray vehicle image is equalized and stretched. Then, the gray image is denoised by median filter method. Finally, the gray image is binarized by global threshold iteration method based on pixel value.For the license plate location, firstly, the binarized vehicle image is processed by Sobel edge detection operator. Then, the binarized vehicle image is scanned from the bottom to the top and the candidate area of the license plate is roughly located and partitioned. Finally, the candidate area of the license plate is processed by mathematical morphology's erosion operation and closing operation. Because the candidate area of the license plate may contain one or more connected areas, the license plate is precisely extracted and partitioned by the method of marking connected areas.For the license plate character segmentation, firstly, because of the license plate's inclined deformation in some degree, the inclined angles of the license plate are obtained by the improved Hough transformation method. And the license plate image is rotated and corrected. On the basis of these operations, the borders and rivets of the license plate are removed. Then, the license plate image is processed by mathematical morphology's dilation operation and opening operation in order to strengthen charaters'information and remove isolated noises. The license plate's characters are segmented by vertical projection method and the license plate's prior knowledge. Finally, the license plate's characters are normalized by bilinear interpolation method. The algorithms in this thesis are implemented in Visual C++ 2005 development environment. The experimental results show that the license plate can be accurately located and the license plate's charaters can be accurately segmented by these algorithms. And the algorithms are robust and real-time.