Target tracking method based on internal clipping and multi-layer feature information fusion

Abstract

The invention belongs to the technical field of computer vision, and discloses a target tracking method based on internal clipping and multi-layer feature information fusion, which comprises the following steps: acquiring a video sequence data set, and forming a training set according to the video sequence data set; constructing a twin network, wherein a basic backbone of the twin network adopts acombination of a ResNet18 feature extraction network improved by an internal clipping unit and an RPN network; based on the training set, training a twin network, and obtaining a training convergenttwin network model; and performing online tracking by using the twin network model. The method solves the problems that a deep learning tracking method in the prior art is poor in feature expression ability and model generalization ability, and a tracker cannot effectively cope with severe space-time change of target appearance.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to an object tracking method based on internal cropping and multi-layer feature information fusion. Background technique [0002] Object tracking technology is an important technical means to extract key information of video, which aims to obtain the position of the object of interest in the video sequence, so as to extract the trajectory of the object. This topic is an important basic topic in the field of computer vision. On this basis, deeper analysis can be carried out, such as abnormal behavior recognition, pedestrian re-identification based on video sequences, etc. [0003] The current mainstream target tracking algorithms include correlation filtering algorithms and deep learning algorithms. Correlation filtering algorithms introduce the concept of signal correlation in signal processing. Representative algorithms include MOSSE, KCF, BACF, etc. Among them, the propo...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a target tracking method based on internal clipping and multi-layer feature information fusion, which solves the problem that the deep learning tracking method in the prior art has weak feature expression ability and model generalization ability, and the tracker cannot effectively deal with the target. The problem of apparent drastic spatio-temporal variation

Images