A Robust Video Target Object Tracking Method

Abstract

The invention discloses a robust video target object tracking method, comprising: sampling and acquiring candidate target particles of the current frame based on the principle of particle filter importance sampling; filtering out divorced particles based on the cosine similarity between the candidate target particles and the previous tracking results ; The linear representation of the screened candidate target particles based on the template dictionary; the low-rank sparse mathematical model of the linear coefficient matrix; the coefficient matrix is ​​solved based on the inexact Lagrangian multiplier optimization algorithm; the best candidate target is selected according to the discriminant function As a result of tracking; the template dictionary is updated online. The robustness of the method of the invention ensures that the method can cope with various challenging factors and accurately track the target object in the video. Overcome inaccurate tracking or even drift of the target object in the video due to illumination changes, scale changes, occlusion, deformation, motion blur, fast motion, rotation, background clutter, low resolution, etc.

Description

technical field [0001] The present invention relates to the field of computer vision, in particular to a method that can cope with various challenges such as illumination changes, scale changes, occlusion, deformation, motion blur, fast motion, rotation, background clutter, and low resolution of target objects in videos. Robust video object tracking method for continuous and accurate object tracking. Background technique [0002] Video object tracking is an important link in computer vision system, which integrates the theoretical knowledge of image processing, stochastic process, artificial intelligence and other fields. Object tracking is to use the color, texture, edge, outline and other characteristics of the specified target to detect the target, use the tracking algorithm to estimate the current state of the target's motion, and predict the state of the next moment, so as to realize the specified target. Accurate, stable, and fast tracking, obtain the motion parameter...

AI Technical Summary

Problems solved by technology

[0004] In view of the important application value of video object tracking technology, academia and companies have conducted extensive research on it, with a research and development history of more than 20 years, but there is still no good tracking algorithm model that can accurately track objects in real time. track

The main reason is that there are various challenges in tracking objects such as lighting changes, scale changes, occlusions, non-rigid object deformations, motion blur, fast motion, rotation, objects leaving the view, background clutter, low resolution, etc., which seriously affect the Tracking accuracy and design algorithm complexity

Images