Weighting local multi-task sparse tracking method with robustness

Abstract

The invention discloses a weighting local multi-task sparse tracking method with robustness, so as to improve the accuracy rate of tracking a target object. The method is as follows: target templates are initialized, and block processing is performed on each target template, thereby obtaining a template dictionary; candidate target particles of a current frame are obtained based on a Gaussian random sampling principle, and block processing is also performed on the candidate target particles; the candidate target particles after block processing is linearly expressed based on the template dictionary; stray particles are captured based on a local multi-task sparse tracking model; a globally optimal solution of the tracking model is obtained based on a near-end gradient acceleration method; a tracking result is determine according to observation likelihood; and the template dictionary is updated on line. The weighting local multi-task sparse tracking method with robustness can effectively prevent a phenomenon that tracking is inaccurate or even drifts from being caused by illumination variation, scale variation, sheltering, deformation, rapid movement, rotation and background clutter of a target object in a video, thereby achieving the purpose of accurately and continuously tracking the target object in the video.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a robust weighted local multi-task sparse tracking method. Background technique [0002] Visual tracking is an important link in the computer vision system, which integrates the theoretical knowledge of image processing, stochastic process, artificial intelligence and other fields. Visual tracking is to use the color, texture, edge and outline of the specified target object to detect the target object, use the tracking algorithm to estimate the current motion state of the target object, and predict the state at the next moment, so as to realize the specified Accurate, continuous and fast tracking of target objects. The key to visual tracking is to establish the corresponding relationship between the candidate target in each frame image and the pre-determined tracking target. [0003] Visual tracking technology has broad application prospects and huge market demand. At present, it...

AI Technical Summary

Problems solved by technology

In view of the important application value of visual tracking technology, both academia and industry have conducted extensive research on it, but due to the existence of complex scenes such as illumination changes, scale changes, occlusion, deformation, fast motion, rotation, and background clutter, time So far, there is still no tracking method that can meet the above-mentioned challenges

Images