| License Plate Recognition system has a wide range of applications. It is a hot point of research in computer vision, image processing and pattern recognition. A whole License plate recognition system generally consists of four components: license plate location, license plate inclination correction, character segmentation and character recognition.Some problems still existed in current working systems.In order to solve problems mentioned above, this paper has completed the following work:In the plate location part, it applies two methods, the first method, aiming at a static image in the fixed location, improves the traditional algorithm which combines edge detection and morphological transformation; The second method applies horizontal scanning combining vertical projection and cross-column scanning to find plate boundary.For inclination correction part, puts forward edge detection based on Canny operator combined with the Probability Hough Transform method todetect inclination angle of the rotating plate and correct it effectively. For character segmentation, it applies coarse segmentation and secondary segmentation method based on priori knowledge.For character recognition, the recognition system is divided into four different sub-systems. It improves template matching method to recognize numbers and letters respectively; puts forward the method which combines template matching and Bayesian classification based on binary data to recognize mixed sub-system of numbers and letters; applies image distance matching technology to recognize Chinese characters.On professional software development tools, adopts Visual C++ with OpenCV library to develop License Plate Recognition System. Experiments show that the system designed in this paper in terms of positioning, inclination correction, segmentation and recognition can attain satisfactory effect, and it has a certain practicality. |
PDF Full Text Request