Unsupervised pedestrian re-identification method based on pixel and feature transfer fusion

Abstract

The invention belongs to the technical field of pattern recognition, particularly relates to an unsupervised pedestrian re-identification method, system and device based on pixel transfer and featuretransfer fusion, and aims to solve the problems of low retrieval precision and poor recognition accuracy of an existing cross-data-set unsupervised pedestrian re-recognition method. The method of thesystem comprises the following steps: acquiring a to-be-identified pedestrian image as an input image; based on the input image, obtaining a pedestrian re-identification result through a pre-trained pedestrian re-identification model, wherein the pedestrian re-identification model is constructed based on a residual network. According to the invention, the retrieval precision and the recognition accuracy of the existing cross-data-set unsupervised pedestrian re-recognition method are improved.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition, and in particular relates to an unsupervised pedestrian re-identification method, system and device based on fusion of pixel and feature transfer. Background technique [0002] The goal of person re-identification is to match pedestrian images from different cameras. For any queried pedestrian image in a camera, it is necessary to find all pedestrian images with the same identity in other cameras. Pedestrian re-identification technology is widely used in video surveillance, smart city, public safety and other fields. However, person re-identification remains a challenging problem due to the drastic intra-class variation caused by view, pose, lighting, occlusion, and small inter-class similarity. [0003] Traditional person re-identification methods can be divided into feature learning based methods and metric learning based methods. However, these methods suffer from handcrafted fea...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above-mentioned problems in the prior art, that is, in order to solve the problems of low retrieval accuracy and poor recognition accuracy of existing cross-dataset unsupervised pedestrian re-identification methods, the first aspect of the present invention proposes a method based on pixel transfer and An unsupervised person re-identification method based on feature transfer, which includes:

Images