Dynamic face recognition system and recognition method based on stereoscopic video stream

Abstract

The invention discloses a dynamic face recognition system and recognition method based on a stereoscopic video stream. The system comprises a video stream, a face detection module, a face tracking module, a face clustering module and a face recognition module. An output end of the video stream is electrically connected with the input end of a face detection module and the input end of a face tracking module. An output end of the face detection module is electrically connected with the input end of the face tracking module, and the output end of the face tracking module is electrically connected with the input end of a face clustering module. According to the invention, the invention has the following advantages, a multi-label classification and multi-task identification method is adopted; key point information and face attribute information of the face are identified while the face is detected; compared with the prior art, the method has the advantages that relevant information of face attributes is recognized while face detection is carried out, the calculation time of attribute recognition is saved, a clustering analysis method based on a tracking result is provided, face images of non-current id in the tracking result are removed, manual deletion and selection work is reduced, and later face data collection work is facilitated.

Description

Technical field [0001] The invention relates to the technical field of face recognition, in particular to a dynamic face recognition system based on a stereo video stream. Background technique [0002] The research of face recognition system began in the 1960s. After the 1980s, with the development of computer technology and optical imaging technology, it has been improved, and the initial stage of application was in the late 1990s, and it was based on the United States, Germany and Japan. The key to the realization of technology; the key to the success of the face recognition system is whether it has cutting-edge core algorithms, and the recognition results have a practical recognition rate and recognition speed; the "face recognition system" integrates artificial intelligence, machine recognition, and machine learning , Model theory, expert system, video image processing and other professional technologies, and need to combine the theory and realization of intermediate value pr...

AI Technical Summary

Problems solved by technology

[0003] The existing dynamic face recognition system for video streaming mainly includes face detection, face tracking, image quality judgment, feature extraction, feature comparison and other modules. At present, the face detection algorithm only extracts the face position information in the picture. Face attribute information (whether it is blurry, whether it is wearing sunglasses, whether it is blocked, etc.) is not identified during the detection process, but a separate quality determination algorithm is used to identify face information, which will cause repeated calculations and increase the time consumption of the algorithm. When faces are occluded or alternated, the tracking algorithm will fail, resulting in different people with the same ID appearing in the tracking results. Before feature extraction, use the quality determination algorithm to find a picture with the best picture quality in the face tracking results Feature extraction is carried out, but the picture does not fuse the information of the front and rear frames, which will affect the accuracy of recognition

Images