Group pedestrian re-identification system based on hybrid attention mechanism and terminal

Abstract

The invention provides a group pedestrian re-identification system based on a hybrid attention mechanism. The group pedestrian re-identification system comprises a backbone model module, a hybrid attention model module, a feature extraction module, a feature fitting module and a feature evaluation module, wherein a deep convolutional neural network backbone model is used to carry out preliminary feature extraction on the group image; the hybrid attention mechanism model is used for further extracting the preliminarily extracted features; and the features of the mixed attention model are compared and evaluated by using a least square residual distance. Meanwhile, the invention provides a terminal for operating the group pedestrian re-identification system. According to the group pedestrianre-identification system, a mixed attention model including spatial attention and channel attention is utilized, so that the network is enabled to pay more attention to the key areas and features of the group images; a novel least square residual distance based on the least square algorithm is proposed, so that the measurement between the group image pairs is better learned, and various challengesbrought by the group pedestrian images can be well adapted; and the group pedestrian re-identification system has good diversity and universal applicability.

Description

technical field [0001] The invention belongs to the technical field of computer vision, specifically a group pedestrian re-identification system and terminal based on a mixed attention mechanism, and relates to group-focused pedestrian re-identification under non-overlapping monitoring cameras. Background technique [0002] In recent years, people have paid more and more attention to public safety issues, and the degree of attention to safety issues has risen to a new height. The video surveillance network in the city is becoming more and more perfect, and video surveillance is spread all over the city. The quality and quality of video surveillance data The number has greatly increased. In all parts of the video surveillance network, considering the flexibility and variability of pedestrian activities and the significance of monitoring pedestrians for protecting personnel safety and criminal investigation and solving crimes, pedestrians are one of the key objects of surveill...

AI Technical Summary

Problems solved by technology

Some works extract subsets in the group and use multi-level matching algorithm to iteratively match, which can fully extract the characteristics of the group and can deal with the change of the number of people in the group (see Xiao, Hao, et al. "Group Re-Identification: Leveraging and Integrating Multi-Grain Information."2018 ACMMultimedia Conference on Multimedia Conference.ACM,2018.), but the matching process is very time-consuming and not suitable for large data sets and real-world scenarios

Images