Multi-camera high-precision pedestrian re-identification method for supervised scenes in same camera

Abstract

The invention discloses a multi-camera high-precision pedestrian re-identification method for supervised scenes in the same camera. The method comprises the steps of photograhing by adopting multiplecameras in the same pedestrian scene, selecting a basic network model, modifying after pre-training, acquiring a pedestrian picture set to be trained, establishing pedestrian memory features for eachcamera, and initializing the pedestrian memory features; based on an existing to-be-trained pedestrian picture set, performing training optimization and supervision in the same camera stage on the basic network model; obtaining a pedestrian pseudo-label by combining the trained pedestrian memory features with a clustering method, and performing fine tuning training on the basic network model by using the pedestrian pseudo-label; and performing cross-camera pedestrian re-identification application on the basic network model obtained by training. According to the method, the recognition performance can be effectively improved only by marking the pictures in the same camera in the scene, the re-recognition accuracy equivalent to that in a fully-supervised scene is achieved, and the pedestrianre-recognition accuracy equivalent to that in the fully-supervised scene is achieved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a multi-camera high-precision pedestrian re-identification method for supervised scenes in the same camera. Background technique [0002] The problem to be solved in pedestrian re-identification is to match the same pedestrian between different cameras; due to the many applications of pedestrian re-identification in security, surveillance, criminal investigation, etc., this task has attracted industry and academic circles in recent years. extensive research and attention in the field. [0003] Although the pedestrian re-identification task has achieved great development, the high performance of the current pedestrian re-identification model and method depends on a large amount of labeled data; Labeling is very expensive and expensive, which limits the application of pedestrian re-identification technology in actual production and life. [0004] In the proces...

AI Technical Summary

Problems solved by technology

[0007] The existing pedestrian re-identification methods for supervised scenes in the same camera, the main problem is that they do not make full use of the known in-camera annotation information to design efficient re-identification models and promote the effective mining of cross-camera pedestrian related information , so it needs to be improved

Images