Matching method and device for pedestrian re-identification, intelligent terminal and storage medium

Abstract

The invention discloses a matching method and device for pedestrian re-identification, an intelligent terminal and a storage medium, and the method comprises the steps: extracting foreground images containing pedestrians in an image library image and a retrieved image, learning the similarity score of each part of the pedestrian in the image library image and each part of the pedestrian in the retrieved image through a trained metric learning model, and obtaining a matching result according to the similarity score of each part; according to the embodiment of the invention, the measurement learning model is obtained based on training of various clothes, various postures and various samples with shielded parts, so that the robustness, the discrimination and the universality of the measurement learned by the model are better, and the accuracy is higher in the matching of long-time pedestrian re-identification in the open world.

Description

technical field [0001] The invention relates to the technical field of video surveillance, and in particular, to a matching method, device, intelligent terminal and storage medium for pedestrian re-identification. Background technique [0002] In recent years, domain-generic and open-world metrics have faced several challenges in the open world. While these metrics are learned from matches from ordinary people, these metrics are learned under naive assumptions, such as learning metrics for small-scale networks with only a few pairs of cameras. Furthermore, these metrics only consider short-term re-identification problems, where images are captured for a limited time, and clothing changes are not considered. In the real world, the same pedestrian can be observed in multiple camera network domains, and, at different time intervals, the pedestrian can be presented with different clothes in different camera network domains; further, the pedestrian can be seen in real-time May ...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to provide a matching method, device, intelligent terminal and storage medium for pedestrian re-identification in view of the above-mentioned defects of the prior art, aiming at solving the problem of only considering well-tailored pedestrians in the prior art. The image of the person in the middle of the image is used as input. The robustness, discrimination and versatility of the measurement obtained by the existing technology are not good, which leads to the problem that the matching accuracy of the open world pedestrian re-identification is not high.

Images