Target specific response attention target tracking method based on twin network

Abstract

The invention discloses a target specific response attention target tracking method based on a twin network, and relates to the computer vision technology. The method aims at overcoming the defect that an original target tracking method based on a twin network is not robust enough in complex tracking scenes such as rapid movement, shielding, rotation and background disorder of a target. The invention provides a target specific response attention target tracking method based on a twin network. The proposed target response attention module effectively weakens the influence of noise information on the tracking performance in the tracking process; meanwhile, the feature information having discrimination for the appearance change of the target object is enhanced, so that the target response graph generated by the twin network can be used for target position prediction, and therefore, more robust tracking performance can be realized. The method comprises five main parts: CNN feature extraction; generating a response graph through channel-by-channel cross-correlation; generating a weight by using an attention network, and weighting each channel response graph; and finally, determining a target position on the response graph, and providing a training method of the model.

Description

technical field [0001] The invention relates to computer vision technology, in particular to a target tracking method based on twin network-based target specific response attention. Background technique [0002] Object tracking, as a fundamental task in computer vision, has a wide range of applications in areas such as video surveillance, vehicle navigation, and augmented reality. Target tracking is to select an object of interest in the first frame of a given video sequence, and predict the position of the target in subsequent frames through computer vision algorithms. In recent years, twin-based tracking methods have attracted the attention of researchers because they can ensure tracking accuracy and real-time speed, but the performance of related algorithms is vulnerable to: fast movement of the target or camera, changes in the appearance of the target, The impact of some unavoidable situations in reality such as background clutter. [0003] In deep learning, the attent...

AI Technical Summary

Problems solved by technology

In recent years, twin-based tracking methods have attracted the attention of researchers because they can ensure tracking accuracy and real-time speed, but the performance of related algorithms is vulnerable to: fast movement of the target or camera, changes in the appearance of the target, The influence of some unavoidable situations in reality such as background clutter

Images