Face Feature Extraction Based On Agents With Multi-camera System

Face recognition technology has always been an important research topic in the field of current pattern recognition and artificial intelligence, its research across the image processing, pattern recognition, computer vision and artificial neural network and other subjects. With the rapid development of computer technology, the increasing needs of commercial and military applications, the research of the face recognition will be more and more depth, it's more likely to be applied to all aspects of daily production and life.Face feature extraction is a complex and challenging issue in face recognition, it has important affect on recognition rate. Because traditional methods of face feature extraction usually affected by light, gaze of the face and other complex environment, the recognition rate may not accurate. In recent research, many robust face feature extraction methods have been proposed, while most of them are improved from matrix algorithm aspect to improve the recognition rate. The research purpose of the paper is to solve the perception problem by multi camera system, and extract face feature with a system of multi camera managed by an Agent framework. The work of the paper includes:Firstly, we have a deep study of the traditional face feature extraction method, especially the ASM method, and realize the classical ASM algorithm, through the experience of two face feature extraction, we summary the virtues and defects of this algorithm.Secondly, propose a method of multi camera system which can solve the problem of lack of face information caused by angles and gaze. We use several cameras to perceive human face environment from different angles, which can enrich the face information, and lay a foundation of face feature extraction.Thirdly, a novel approach was proposed to extract face feature base on Agents with multi-camera system. According to the ability of cooperation and intelligence of the Agent, we can give each camera an Agent, each Agent tries to get the pose parameters of face feature from the environment, also the Agent will react to environment, which will compute the facial pose parameters. Those Agents also compete with each other, in order to become the best facial object. In this way the face feature can be better extracted with the Agent system.Finally, We collect a couple of face pictures from the multi camera system, and have a face recognition test with the ORL face database which recognition rate experiment shows that the method can better extract the face feature. From the experience we also solve the error identification problem caused by face angles, Due to the parallelism of the Agent platform, the processing time also have a big advantage.