Target tracking method based on TLD algorithm

Abstract

The invention discloses a target tracking method based on a TLD algorithm. The target tracking method comprises the following steps: (1) in the initial frame, a tracking window designated by a user togenerate a positive and negative sample initialization detection module and form a target model; 2) in the tracking process, the detection module and the tracking module work independently: the detection module scans the current frame image to obtain image blocks and sequentially passes through a variance classifier, a merging classifier and a nearest neighbor classifier; the tracking module adopts the median optical flow method with the failure detection mechanism to predict the target position of the current frame; (3) the integrating module integrates the detection module and the trackingmodule to output the tracking result; (4) new positive and negative samples are generated in the updated position and weighted by the current tracking result confidence and the detection module is updated by the weighted positive and negative samples; and (5) circulating the steps (2)-(3) until the end of tracking. The comparative experiments performed on the open dataset show that the method canimprove the robustness and the accuracy of tracking to some extent.

Description

technical field [0001] The invention relates to a target tracking method based on a TLD algorithm, which is characterized in that the target model is weighted by using the current frame tracking result confidence, so as to obtain a target model with more representation ability. It is applied to computer vision, target detection, target tracking, etc., and belongs to the field of target tracking in computer vision. Background technique [0002] The TLD tracking algorithm is a single-target long-term tracking algorithm proposed by Zdenka Kalal. The algorithm consists of three modules: tracking module, detection module and learning module. The pure tracking algorithm is difficult to correct the drift error of the tracking and will keep accumulating the tracking error, and once the target disappears from the field of view, the tracking will inevitably fail. Simple detection algorithms require a large number of samples for offline supervised training, and cannot be applied to t...

AI Technical Summary

Problems solved by technology

The rougher the target model, the more errors produced by P-N learning will reduce the representation ability of the target model

Images