Video Streaming Real-Time Tracking Of Human Faces

Video tracking is one of the hot field of computer vision. The main video stream herein face tracking key technology to improve a face detection algorithm, a new real-time video streaming human face tracking algorithm. It has broad application prospects in public safety, video surveillance, property security and other fields.Feature extraction is the basis of video tracking. Based on the classic local texture features, namely Local Binary Pattern(Local Binary Pattern, LBP) and certain of its variants, as well as local ternary mode(Local Ternary Patterns, LTP) and improved methods, focusing on the illumination brightness and attitude change, especially when the light dims, the traditional local texture feature extraction methods are sensitive to noise, lack of robustness of deterioration, is presented based on CS-SILTP( based on scale invariant local gradient information ternary mode) improvements algorithms, and experimentally verified the effectiveness of the improved algorithm and robustness.TLD based on the classic real-time tracking algorithm and its improved algorithm, combined with the proposed CS-SILTP improved algorithm, real-time tracking algorithm based on the probability of anastomosis. Unlike traditional video tracking algorithm for tracking a first frame is detected target, several improved algorithm in the video stream is sampled once every frame image, and the comparison with the threshold value, to find that a higher reliability frame image of this frame as a two-way track points, build tracking path. Comparison with existing tracking algorithm experiments show that the new proposed algorithm has significant superior efficacy and robustness.A large number of comparative experiments show that the proposed improvements based on CS-SILTP algorithms and real-time tracking algorithm based on the probability of consistent, in the light of changes in attitude and changes in circumstances, effective real-time detection, target tracking designees face, showing better tracking performance and robustness.