Video portrait tracking method based on human face identification in complex scenario

Abstract

The invention relates to a video portrait tracking method based on human face identification in a complex scenario. The method comprises the steps that 1) acquired video data are decoded to acquire a series of video frame images; 2) human face detection and object tracking methods are used to process a series of acquired video frame images to acquire all possible human face trajectories; 3) the similarity between each human face trajectory acquired in step 2) and a subsequent human face approaching trajectory is determined, and human face trajectories with the similarity higher than a set threshold are merged; and 4) a merged human face trajectory is output. Compared with the prior art, the method provided by the invention has the advantages of fast processing speed, high detection precision, low false alarm rate and the like. The problems of human face trajectory disruption and confusion, which are caused by human head turning and blocking, can be solved.

Description

technical field [0001] The invention relates to the technical field of video processing, in particular to a video portrait tracking method based on face recognition in complex scenes. Background technique [0002] Face tracking is the process of determining the trajectory and size changes of a certain face in a video or image sequence. For a long time, face tracking has been of great significance in the fields of image analysis and recognition, image monitoring and retrieval, such as video MMS in mobile phone MMS, man-machine interface, authority control, intelligent monitoring system, etc. The accuracy, precision and recklessness of tracking The stickiness problem has always been the main concern of the industry, and many effective algorithms have appeared one after another. The current mainstream video portrait tracking algorithms mainly include the following: [0003] a) Method based on face detection: Face detection is performed on each frame image in the video, and th...

AI Technical Summary

Problems solved by technology

[0005] The above scheme has a better effect when targeting simple scenes (clear faces, mainly frontal faces, no occlusion), but the above scheme does not make full use of video information. For complex scenes (frequent occlusion, face rotation range Large) when the face track is prone to interruption, face track confusion, etc., it is easy to cause false positives and poor performance