An anti-occlusion target tracking method based on KCF and particle filter

Abstract

The invention relates to an anti-occlusion target tracking method based on KCF and particle filter. KCF method and particle filter method are fused to make full use of the advantages of particle filter method, that is, the negative effect of occlusion problem can be responded positively, and the tracking instability of KCF method in the occlusion problem of target tracking can be overcome. The invention utilizes the fast and accurate characteristic of KCF target tracking under the condition of no occlusion, combines with the characteristic of particle filter which can realize good prediction target position under the condition of no occlusion, and proposes a fusion method on the basis of not increasing storage cost and time cost. The multi-scale detection of the target is carried out by adopting the thread pool technology. In the invention, the comparison and combination of the three scales make the obtained scales change more widely and the output scales are selected more reasonably.Because of the pertinence of target sampling, the requirement of hardware storage unit for tracking system is reduced, which makes the tracking more smooth and stable, and the anti-occlusion is more prominent.

Description

technical field [0001] The invention belongs to the technical field of computer vision target tracking, and in particular relates to an anti-occlusion target tracking method which combines KCF and particle filter. Background technique [0002] With the development of computer vision and artificial intelligence, object detection and tracking has been paid more and more attention, and has become a key technology for unmanned driving, drones and intelligent transportation. However, the current single method of target tracking cannot solve the problems of tracking speed, illumination scale change and complex background occlusion. [0003] KCF is a discriminative tracking method, which generally trains a target detector during the tracking process, uses the target detector to detect whether the predicted position of the next frame is a target, and then uses the new detection result to update the training set and then update the target Detector. The advantage of this method is t...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide an anti-occlusion target tracking method that combines KCF and particle filter to solve the problem that the above-mentioned KCF method cannot effectively deal with occlusion when performing target tracking.

Images