Maximum-particle-size structure descriptor-based pedestrian re-identification method

Abstract

The invention relates to a maximum-particle-size structure descriptor-based pedestrian re-identification method. The method comprises the following steps of S1, obtaining colorful pedestrian images in an image set, and processing the pedestrian images by using a Gabor filter to obtain a plurality of dimensional images; S2, obtaining color difference histograms (CDH) of the dimensional images, and extracting LMCC descriptors; S3, extracting LOMO descriptors; S4, performing measure learning by using an LDA algorithm to obtain optimal sub-space projection of a feature space; and S5, inputting to-be-identified pedestrian images, and calculating similarity measure distances between the to-be-identified pedestrian images and the pedestrian images in the image set to obtain an identification result. Compared with the prior art, the method has relatively good robustness for change of factors such as illumination, rotation, translation and the like, can be used for extracting essential features of the images, has good pedestrian identification performance, and is not sensitive to change of illumination, view angles, shielding and the like.

Description

technical field [0001] The invention relates to a structure feature extraction and metric learning technology in pedestrian re-identification, in particular to a pedestrian re-identification method of maximum granularity structure descriptor. Background technique [0002] Pedestrian re-identification refers to the problem of identifying and matching pedestrians under different cameras in a surveillance network composed of multiple cameras. It provides key assistance to the research of identifying pedestrians and analyzing pedestrian behavior, and has developed into an important part of the field of intelligent monitoring. [0003] Pedestrian re-identification methods are mainly divided into two categories: 1) Pedestrian re-identification methods based on feature representation; 2) Methods based on metric learning. Most of these methods focus on finding a robust feature to describe pedestrians, such as: color histogram, co-occurrence matrix, feature axis, maximum stable extr...

AI Technical Summary

Problems solved by technology

[0005] Patent CN104992142A proposes a pedestrian recognition method based on the combination of deep learning and attribute learning, which can describe pedestrian features from a higher semantic level. However, the training model is too complicated and limited by the selection of pedestrian attributes

Furthermore, due to the influence of various factors such as illumination changes, posture, viewing angle, occlusion, image resolution, etc., this makes the performance of pedestrian re-identification in the intelligent analysis of surveillance video still poor.

Images