Pedestrian re-identification method and device

Abstract

The invention provides a pedestrian re-identification method and device. The pedestrian re-identification method includes the steps: measuring the similar distance of each test sample pair in a test set one by one based on a metric space; and sorting the similar distance measurement results of all test sample pairs in the test set to obtain a distance measurement matrix. The pedestrian re-identification method obtains a group of stable combined metric space through iterative learning, calculates the combined similar distance of the test sample pairs in the multi-metric space, overcomes the over fitting problem of an existing metric learning method in the high-dimensional small sample problem, and has higher robustness for the complex feature change in the pedestrian re-identification problem.

Description

technical field [0001] The present invention relates to the technical field of information processing, and more specifically, to a pedestrian re-identification method and device. Background technique [0002] With the progress of society and the rapid development of science and technology, the intelligent monitoring system based on computer vision technology has been greatly promoted and applied in the fields of communication, transportation, security and so on. It plays an important role in preventing and combating crimes and protecting citizens' personal and property safety. effect. [0003] In recent years, with the intelligentization, diversification, and popularization of computer vision application technology, various needs that meet the actual needs of life have been proposed, and the problem of pedestrian re-identification is one of the important hot topics. The goal of person re-identification is to match pedestrian images across views in a non-overlapping surveill...

AI Technical Summary

Problems solved by technology

[0006] However, there are certain problems in the above three types of methods: the method based on the appearance feature model, in the actual environment, due to the drastic changes in shadow, illumination, posture, viewing angle and background, it is difficult to deal with pedestrian features based on color and texture. In the re-identification problem, the visual characteristics of pedestrians under different cameras change, causing the distance between positive samples to increase

In addition, due to the similarity of people's clothing, pedestrian re-identification is also faced with the problem that the distance between positive and negative samples is too small

Although through the efforts of researchers, more effective new feature models have been continuously proposed, but it is still impossible to find a robust and powerful feature to meet the task requirements of pedestrian re-identification.

The method based on metric learning is limited by the instability of sample features and the defect of expression ability, and the algorithm recognition rate is still limited

In addition, the problem of person re-identification is a typical small-sample problem, and the samples have high-dimensional features. The above-mentioned metric learning method has strong constraints, which leads to overfitting of the learned metric space to the training samples, and the general performance on the test samples.

The pedestrian recognition method based on the deep learning method requires a large number of samples, and the training time is long and takes up a lot of resources.

Images