A Target Tracking Method Combining Spatial-Temporal Context Information

Abstract

The invention belongs to the field of pattern recognition and computer vision, and discloses a target tracking method combined with spatio-temporal context information, including: using the first frame of pictures to train an initial strong classifier, and learning the spatio-temporal context model required for next frame tracking; When a new frame arrives, use the trained strong classifier to evaluate several blocks in the search area to obtain the first confidence matrix; then integrate the spatio-temporal context information to obtain the confidence map function, and use this function to obtain each block in the search area confidence value, to obtain the second confidence matrix; finally according to the corresponding weight linear combination to obtain the final confidence matrix, find the block with the largest confidence value in the final confidence matrix is ​​the target to be tracked; the present invention combines the spatiotemporal context information of the target In the online Boosting algorithm, fast robust tracking can be achieved.

Description

technical field [0001] The invention belongs to the technical fields of pattern recognition and computer vision, and in particular relates to a target tracking method combined with spatio-temporal context information. Background technique [0002] Moving object tracking is one of the important research directions in the field of computer vision, and has important applications in human-computer interaction, intelligent monitoring, medical imaging and other fields. Tracking algorithms have made great progress in recent years, but how to effectively solve the problem of tracking drift caused by occlusion, fast motion, illumination changes, background clutter and other factors is still a very challenging problem. [0003] The online Boosting algorithm, when a new frame arrives, uses a strong classifier to classify the background and target in the picture to obtain the target area, but when the target is occluded, it will use the occluded feature to update the feature pool, Poll...

AI Technical Summary

Problems solved by technology

[0007] In view of the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a target tracking method combined with spatio-temporal context information, which can solve the tracking drift that occurs when the target area is partially occluded or the target scale changes greatly in the prior art. question

Images