Target tracking method based on cost-sensitive structured SVM

Abstract

The invention discloses a target tracking method based on a cost-sensitive structured SVM. The method comprises the following steps: firstly, establishing a cost-sensitive structured SVM model for target tracking, and converting the cost-sensitive structured SVM model into a dual problem according to a Lagrange multiplier method; then solving the cost-sensitive structured SVM model by adopting a dual coordinate descent principle, and estimating the state of the target; and finally, performing evaluation by adopting a multi-scale target tracking method, and selecting the maximum response as a tracking result. According to the target tracking method based on the structured SVM, the problem that ease pair and hard pair are unbalanced in an existing target tracking method based on the structured SVM is solved, and the accuracy and success rate of a target tracking algorithm based on the structured SVM are improved.

Description

technical field

[0001] The invention relates to the technical field of computer vision target tracking, in particular to a target tracking method based on cost-sensitive structured SVM. Background technique

[0002] Object tracking is a basic research topic in the field of computer vision and an important technology in video analysis, whose goal is to estimate the state of the object using video data. Target tracking has important application value in civilian fields such as video surveillance, vehicle navigation, human-computer interaction, intelligent transportation, motion analysis, and attitude estimation, as well as in military fields such as visual guidance, target positioning, and fire control. In recent years, although target tracking has made great progress, it still faces many problems such as complex background, target change and fast movement, and it is still a hot issue in the field of computer vision.

[0003] At present, researchers in the field of object tra...

Images