Embedded Human Face Recognition System And Research Of Algorithms Of Charactersegmentation For License Plate

With the rapid development of embedded technology, embedded devices are more and more favored and used widely in all walks of life with their unique features such as flexibility, application specific and portability. Meanwhile, The embedded devices are able to deal with complex affairs with the processing capacity being more and more strong. Therefore the application based on embedded platform will be very widely. As a result it’s a very important demand of the information processing intellectualization that transplant and promote the existing cumbersome systems based on PC platform to embedded platform. This thesis chooses two important technique of the field of intelligent security monitoring, human face recognition and license plate recognition as the subject, and researches the transplantation and optimization of them based on the DM642platform.In this thesis the main results obtained are as follows:1. This thesis studies the algorithms of face recognition and license plate recognition, and successfully transplants the algorithm of face detection and recognition to DM642platform according to the structural characteristics of the platform. This thesis also improves the accuracy of face location by improving the algorithms of face detection and eye location of the system. Also the multi-level optimization was conducted on the system of face recognition according to the characteristics of DM642platform. Finally with the processing time of each frame decreasing to about200ms from2000ms, the system basically reaches the aim of real-time processing. 2. Also this thesis gives the analyzing and exploration to the algorithms of license plate location and the character segmentation, which are important parts of the algorithm of license plate recognition. An improved license plate location and character segmentation algorithm based on the ternary image is proposed by combining the existing algorithm and the license plate detection algorithm based on nearest neighbor chains. And the experimental results show that the proposed algorithm obtains good robustness and adaptability by achieving more than96%accuracy rate of characters segmentation and recognition in different conditions.