Study On Face Detection And Recognition

The technique of face detection and recognition is a fundamental and important subject in the research of Pattern Recognition and Computer Vision. And it plays an important role in many fields, such as content-based image and video retrieval, video surveillance, automatic face recognition and human-computer interface, etc. Based on the Analysis and summary on the related research at home and abroad, this thesis realized the face detection and recognition on computer. What's more, the face detection part was implemented on the DSP platform. At last, side face-based multi-pose face recognition was given deep study so as to solve the hot issue in this field.The main contents of this thesis include five aspects:①A popular algorithm, Adaboost face detection algorithm was researched and realized. Face detection classifier was trained with the help of Adaboost algorithm.②PCA face recognition algorithm was studied and realized. And PCA face recognition system was implemented on Yale face database.③Multi-pose face recognition was deeply studied. Integral projection method was used for face normalization on CMU PIE face database to constitute a new multi-pose face database. Conventional LLR algorithm for multi-face rcognition was given intensive study, and was improved by joining a local constant factor.④A new multi-frontal pose face recognition method based on frontal face compensation algorithm was proposed by introducing a frontal face compensation operator, which was brought from the average frontal face and the average side face. The side face compensated was used for face recognition to reduce the computational complexity and improve the recognition rate.⑤Face detection algorithm on OpenCV was transplanted to DSP development board. And face detection system was realized on hardware platform.