A Pedestrian Re-Identification Method Based on Multi-layer Supervised Network

Abstract

The invention relates to a pedestrian re-identification method based on a multi-level supervision network, which extracts features of different semantic levels from pedestrian images through a multi-level supervision network, and then realizes pedestrian re-identification; the multi-level supervision network includes a multi-layer The deep convolutional neural network is used as the backbone network and multiple classification modules are used as the feature extraction sub-network; the backbone network converts pedestrian images into feature maps of different semantic levels, and each classification module transforms the feature maps of each layer extracted by the backbone network through supervised learning. As a distinguishing feature vector, the feature vectors on all levels are spliced ​​into the final feature vector, and pedestrian re-identification is realized based on the final feature vector. Compared with the prior art, the present invention extracts features of different semantic levels of pedestrian images, improves the distinguishability of features, and uses a semi-separated supervised learning method to improve the stability of the training process and improve the accuracy of the network. The advantages of high re-identification accuracy.

Description

Technical field [0001] The present invention relates to a procedural reintegration method, in particular, to a multi-level supervision network based personnel rereading method. Background technique [0002] The pedestrians in the video reactively recognize an important research topic in computer vision and artificial intelligence. Its task objectives can be briefly outlined as: a given image (or more) to check the image of the pedestrian (Query Image), you need to search for all images of the pedestrian in the existing monitor video set (Gallery images). Pedestrians have a major realistic and value in the fields of intelligent security, urban security, and is a hot spot in recent years. [0003] However, in a realistic scenario, due to the camera shooting angle, shooting distance, the illumination environment, etc., there is a significant visual difference in different video in different video. In addition, the gesture change, occlusion, etc. produced by human movement, further i...

AI Technical Summary

Problems solved by technology

[0013] One of the purposes of the present invention is to solve the problem that the existing pedestrian re-identification technology does not have a high utilization rate of the features of the middle layer of the convolutional network, and improve the differentiation and robustness of the overall features

Images