Pedestrian re-identification method and device, computer equipment and storage medium

Abstract

The invention provides a pedestrian re-identification method and apparatus, computer equipment and a storage medium. The method comprises: obtaining a plurality of candidate posture images and corresponding candidate face images of candidate pedestrians; determining identification information of the candidate pedestrian according to a preset mapping relation between the candidate face image and the identification information; acquiring a first posture image of a target pedestrian on site, and extracting a target feature corresponding to the first posture image by using a feature extraction model; comparing the target feature with a plurality of candidate features corresponding to the plurality of candidate posture images, wherein the candidate features are obtained by performing feature extraction on the candidate posture images through the feature extraction model; and determining a candidate feature consistent with the comparison result of the first feature, and taking the identification information corresponding to the candidate feature consistent with the comparison result as the identification information of the target pedestrian.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a pedestrian re-identification method, device, computer equipment and storage medium. Background technique [0002] Pedestrian re-identification means that on the basis of pre-acquired pedestrian images and their identified identities, when the images of the pedestrians are captured again, the recognition model can be used to extract pedestrian features to confirm their corresponding identity information. However, the current pedestrian re-identification technology mainly has two disadvantages. On the one hand, most of the current pedestrian re-identification methods need to associate different pedestrian images with their names and other identity information through manual labeling, which is poor in automation and is not conducive to timely update of the image database. On the other hand, the existing recognition models are all trained based on labeled data. Due to th...

AI Technical Summary

Problems solved by technology

On the one hand, most of the current pedestrian re-identification methods need to associate different pedestrian images with their names and other identity information through manual labeling, which has a poor degree of automation and is not conducive to the timely update of the image database.

On the other hand, the existing recognition models are all trained based on labeled data. Due to the small amount of labeled data and the single scene, most of the existing recognition models have defects such as low recognition accuracy and poor robustness.

Images