计算机识别几何畸变车牌图像的关键技术研究-计算机技术专业论文.docxVIP

华中科技大学硕士学位论文 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract Therefore, intelligent transportation becomes the hot research in recent years. As a indispensable part in intelligent transportation systems ， license plate recognition application scene also becomes more complicated. Accuracy is particularly important. The existing vehicle plate recognition algorithm is simple and such recognition algorithm requires certain angle and height of the camera installation. Its monitoring perspective is extremely limited. When the camera is tilted, the vehicle plate in the image is geometric distorted，which is hard to be correctly segmented and recognized. So segment and recognition are important problems which must be solved. In this paper, based on the existing research algorithm, knowing general location in the image of the vehicle plate, the geometric distorted vehicle plate recognition algorithm is studied and analysised. The geometric distorted recognition algorithm guarantees the effectiveness of recognition and increases the camera supervising scope and angle. The study mainly focus on the geometric distorted correction, segmentation and an enhancement algorithm for recognition—shape context. The study show that:Using Hough, get the image of the vehicle plate tilt direction and then make the distorted vehicle plate recovery with specific geometric transformation. Put forward a new kind of vehicle plate character segmentation algorithm -based on Gaussian distribution, adaptive template matching segmentation algorithm. This algorithm can overcome the small stain on vehicle plate and certain geometric distortion. Shape context algorithm has robustness for zoom, shift and tilt. Even the geometric distorted vehicle plate still cannot completely recovery after geometric distorted correction, they still can be recognized by the shape context algorithm. Key words: Vehicle plate recognition, Geometric distorted, Character Segmentation, Character recognition 独创性声明 本人声明所呈交的学位论文是我个人