Pedestrian re-identification method based on spatial and temporal features

Abstract

The invention provides a pedestrian re-identification method based on spatial and temporal features. The method includes: firstly, adopting a human body posture estimation method to extract pedestrianskeleton key points; for each frame of image in a video sequence, dividing a human body into a human body main body part and a human body affiliated background part according to the skeleton key points, and designing a dual-channel neural network to extract the apparent characteristics of pedestrians in the image; for the video sequence, obtaining the space-time information of the pedestrian in the video sequence through series superposition of the time information extracted from the adjacent frames and the space information extracted from the spatial image convolution; and finally, combiningthe extracted single-frame image apparent characteristics with the extracted video space-time characteristics, considering the weight influence of the time dimension on characteristic fusion, and distinguishing pedestrian characteristic pairs by using a metric learning method. According to the method, the two-channel neural network is designed to extract the apparent characteristics of the pedestrian in the image, and based on the deep characteristic analysis, the pedestrian re-identification efficiency can be effectively improved, and the high pedestrian re-identification precision can be maintained.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a pedestrian re-identification method based on spatio-temporal features. Background technique [0002] With people's increasing attention to social public safety and the development of video capture technology and large-scale data storage technology, a large number of surveillance cameras are used in shopping malls, parks, schools, hospitals, companies, stadiums, large squares, subway stations and other densely populated areas. Places prone to public safety incidents. It has been difficult for humans to cope with the massive growth of surveillance videos, so the need for re-identification of pedestrians in surveillance videos by computers has arisen. [0003] Pedestrian re-identification is also called pedestrian re-identification, which is to identify pedestrian images taken at different times under cross-camera and cross-scene conditions, so as to judge whether they ...

AI Technical Summary

Problems solved by technology

[0004] Although face recognition technology has developed relatively maturely and has been applied in many scenarios and products, the application of face recognition technology has certain limitations: First, face recognition technology can only be used for people with a human body. Face information, other important information is not fully utilized, such as: clothing, posture, behavior, etc.; secondly, face recognition technology must have a clear frontal face photo when it is applied, that is, the image details are relatively high. These conditions cannot be met in many scenarios, such as: lowering the head and side face, facing away from the camera, blurring the body shape, blocking the hat, etc.

However, due to the variety and complexity of monitoring scenes and the non-cooperative nature of pedestrians, there are still many challenges in the research of pedestrian re-identification, and the recognition accuracy and recognition efficiency need to be further improved.

Images