Multi-feature selection target tracking method based on support vector machine

Abstract

The invention discloses a multi-feature selection target tracking method based on a support vector machine, the steps of which are: acquiring positive and negative samples, extracting various features of the samples; extracting candidate targets, extracting multiple features of the target; and performing kernelization processing on all features ; Train the SVM classifier; calculate the confidence value, identify the candidate target; update the classifier; evaluate the tracking performance. The present invention can solve the problem of poor ability to describe the target by a single feature and overcome the disadvantage of weak classification ability of the classifier trained by a single positive sample, and can improve the accuracy of the target tracker in complex scenes such as large illumination influence and serious occlusion of the target and robustness.

Description

Technical field [0001] The invention belongs to the field of image processing and computer vision, and particularly relates to a multi-feature selection target tracking method based on a support vector machine. Background technique [0002] Target tracking is a key issue in the field of computer vision research, and it can be used in many application fields such as automatic monitoring, robot navigation, and human-computer interaction. For a robust visual tracking algorithm applied to real scenes, it needs to effectively deal with the changes of target appearance and posture, rapid motion, occlusion, lighting effects, background clutter and other challenges. [0003] At present, the mainstream target tracking method is to treat the tracking problem as a classification problem. That is, the target of interest is separated from the background in a specific area. The patent "Visual target tracking method based on multi-feature joint sparse representation" (publication number: CN 103...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problems raised by the above-mentioned background technology, the present invention aims to provide a multi-feature selection target tracking method based on support vector machines, which can solve the problem of poor ability of a single feature to describe the target and overcome the classifier classification ability of single positive sample training Weaker shortcomings, which can improve the accuracy and robustness of the target tracker in complex scenes such as large lighting effects and severe occlusion of the target

Images