Pedestrian re-recognition method based on multi-layer neural network for cross-scene pedestrian recognition

Abstract

The invention discloses a method for re-identifying pedestrians across scenes based on a multi-layer neural network, which comprises the following steps: (1) collecting pedestrian video in a camera under a current scene, and intercepting video frames; (2) feature extraction and dimension reduction of the pedestrian image in the video frame intercepted in the step (1) being carried out, and the training set Xt of the target domain being constituted by using sample pairs to obtain a test set Xo; (3) processing the identified data under the relevant scenarios, and using sample pairs to constitutethe source domain training set Xs; (4) establishing training set X, X= [Xs, Xt]; (5) using X to train the multi-layer neural network model; (6) identifying the sample to be identified in the test setXo according to the model obtained in step 5. The invention selects a multi-layer neural network with complex non-linear mapping as a learning model, and adds the identification data of the relevantscene to the model learning of the new scene by using the transfer learning idea, so that the learning of the new scene is more accurate and effective.

Description

technical field [0001] The present invention relates to the field of computer vision and pattern recognition, in particular to a method for cross-scene pedestrian re-identification based on a multi-layer neural network Background technique [0002] With the popularization of a large number of cameras in public places, the application of pictures and video data based on pedestrians has gradually received more and more attention. One of the important applications is pedestrian re-identification. Pedestrian re-identification refers to the technology of finding the same pedestrian through multi-time segment pedestrian video data collected by non-overlapping cameras. Today, as public safety is increasingly valued, pedestrian re-identification has attracted more and more attention. With the application of pedestrian re-identification in various new fields, a very important technical issue under actual conditions is how to deploy a pedestrian re-identification system in a new scen...

AI Technical Summary

Problems solved by technology

[0004] The main purpose of the present invention is to use the nonlinear and high-precision processing capabilities of the neural network to provide an easy-to-operate and highly reliable cross-scene pedestrian re-identification method based on a multi-layer neural network, focusing on improving the existing method The defect that the recognition accuracy of the established cross-scene pedestrian re-identification model is not high

Images