A Siamese Network Object Tracking Method Based on Channel Attention Update Mechanism

Abstract

The invention discloses a twin network target tracking method based on a channel attention update mechanism, which relates to the field of image processing. The twin network is used to extract features, select foreground templates and background templates, and highlight target template features and foreground template features through an adaptive channel selection mechanism. to activate the effective foreground channel to suppress the background feature channel; generate the preliminary tracking results through the candidate region generation network, trigger the template update mechanism with the confidence decision tracking confidence, generate the update template through the channel attention update network, and use the target template to match Update the template for re-tracking, fuse the re-tracking results, and correct the tracking error. It improves the foreground feature extraction ability of the tracker, improves the ability of the target tracking method to discriminate against background interference, makes up for the disadvantage that it is difficult to deal with complex tracking situations without online template update during the tracking process, avoids tracking drift, and makes the tracking process more efficient. Robust, more accurate tracking accuracy.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a twin network target tracking method based on a channel attention update mechanism. Background technique [0002] Object tracking is the most basic research hotspot in the field of computer vision, and has a wide range of applications in robotics, human-computer interaction, intelligent vehicles, monitoring and other fields. Although a large number of tracking methods have been proposed for various scenarios, robust and accurate visual tracking methods are still difficult to achieve due to factors such as deformation, occlusion, illumination changes, background clutter, and fast motion. [0003] In recent years, Siamese networks have attracted great attention in the tracking community due to their balanced accuracy and speed. By defining object tracking as a matching problem, the Siamese tracking method aims to learn a general similarity function offline from a large number of v...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is: how to solve the problem that the twin candidate region generation network method cannot perform robust tracking and is susceptible to similar objects and background interference due to the lack of template update during the tracking process

Images