Adaptive scale video target tracking method

Abstract

The invention provides an adaptive scale video target tracking method. According to the method, firstly position and scale information of an initial frame target is given and a training sample is acquired by performing cyclic shifting; then the nuclearized ridge regression classifier is solved so as to acquire a filtering template, and a candidate sample is detected to estimate the position of the target; then a scale sample is detected based on the solved target position and the size of the previous frame of target, and the scale of the present frame of target is determined by the maximum response value; and finally the target and scale model parameters are updated online according to the occlusion detection mechanism. The method is adaptive to illumination and attitude change, background interference and other complex scenes, and the target can still be effectively tracked when the target is under severe occlusion and scale change so as to have great tracking accuracy and robustness.

Description

technical field [0001] The invention belongs to the technical field of computer vision and relates to a video target tracking method adaptive to scale changes, which is applied to the tracking of moving targets in the field of intelligent video processing, and is suitable for tracking when the scale of the moving target is constantly changing. Background technique [0002] Moving target tracking is one of the research focuses in the field of computer vision. Its purpose is to estimate the position and trajectory of the target of interest in the video sequence. Wide range of applications. [0003] In recent years, the discriminative-based tracking method has shown good tracking performance. This method introduces the idea of ​​machine learning, regards target tracking as a binary classification problem, and uses the background information of the target and its surroundings to learn a discriminative formula online. The classifier of , evaluates the candidate samples to locali...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to improve the KCF moving target tracking algorithm for the problem that it cannot adapt to target scale changes and template updates, and to provide a scale-adaptive video target tracking method. The size of a frame target is obtained from the scale sample for detection, and the size of the current target is estimated from the scale corresponding to the maximum response value; at the same time, occlusion detection is added in the template update process. When the target is judged to be occluded, the target model containing interference information is weakened. Weighting in Classifier Model Updates for Continuously Robust Tracking of Objects

Images