Target re-identification method and device

Abstract

The invention discloses a target re-identification method and device, computer equipment and a computer readable storage medium. The method comprises the following steps: acquiring an expansion image corresponding to each target image in an image set; based on the graph neural network, obtaining a graph embedding vector of the expansion graph corresponding to each target image; and based on the image embedding vector of the expansion image corresponding to the to-be-identified target image and the image embedding vectors of the expansion images corresponding to other target images, determining a target which is the same as the to-be-identified target, wherein the expansion graph corresponding to any target image at least comprises: a central node taking the target as a center, nodes taking other targets, the distances between which and the target are within a first threshold value, as adjacent nodes, and nodes taking each adjacent point as a center and other targets, the distances between which and each adjacent node are within a second threshold value, as adjacent points, and edges, to which of the corresponding central nodes are pointed by the adjacent nodes. According to the scheme of the invention, the accuracy of target re-identification is improved to a certain extent, and the robustness is high.

Description

technical field [0001] The present invention relates to the technical field of computer vision, in particular to a target re-identification method and device, computer equipment, and a computer-readable storage medium. Background technique [0002] Target re-identification is also called target re-identification. Currently, it mainly focuses on the fields of pedestrian re-identification and vehicle re-identification. Taking person re-identification as an example, it can be defined as searching out the target pedestrian from a large-scale candidate pedestrian database for a given pedestrian in the camera in a non-overlapping video surveillance network. Pedestrian re-identification can quickly locate pedestrian targets of interest in videos generated by a large number of surveillance networks. It is widely used in security and public places to find people. It is an important technology for intelligent video surveillance and has received extensive attention in the field of comp...

AI Technical Summary

Problems solved by technology

However, this technology mainly focuses on extracting sufficiently discriminative individual features for pedestrians, which leads to the fact that pedestrian re-identification algorithms based on individual features are difficult to be effective when the appearance of pedestrians changes greatly due to changes in factors such as illumination, viewing angle, posture, and occlusion. identify the target pedestrian

Images