Target tracking method based on deep convolution feature adaptive integration

Abstract

The invention discloses a target tracking method based on deep convolution feature adaptive integration. The method comprises the steps of extracting deep convolution features; calculating a kernel correlation filter; updating the integrated vector of the current frame by using an integrated vector updating formula; predicting a target position of the current frame image by using a self-adaptive integrated calculation formula; updating the deep convolution feature of the current frame by using a deep convolution feature updating formula; and taking the target center position of the current frame as the center position of the to-be-tracked target when the iteration of the video image sequence containing the to-be-tracked target is ended. By integrating the features, the defect that a tracker in the prior art cannot fully utilize information contained in different channel target features is overcome, so that the position of the target to be tracked is more accurately acquired in the target tracking process, and the accuracy and reliability of target tracking are enhanced.

Description

Technical field [0001] The present invention belongs to the technical field of image processing, computer vision still further relates to a technical field of image processing based on the depth of target tracking and adaptive convolution integration features. The present invention employs an adaptive integration method based on convolution filter associated feature depth, for video surveillance, moving object in the field of health care, intelligent transportation, robot navigation, human-computer interaction, virtual reality tracking. Background technique [0002] The main task is to estimate the trajectory of target tracking an object in the video, i.e., detect moving targets to be tracked in the sequence of video images, and to determine the position of the moving object in each frame image. One of the most popular target tracking method is to detect tracking, detection and tracking of learning is usually a binary classifier to separate the target and the background based on ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to address the shortcomings of the above-mentioned prior art, and propose a target tracking method based on adaptive integration of deep convolution features, which is used to solve the problem that the tracker cannot make full use of the target features of different channels when there are similar disturbances around the target. The tracking information is not stable enough, and the obtained target position is not accurate enough

Images