Human body target re-identification method and system among multiple cameras

Abstract

The invention is suitable for human body target re-identification among multiple cameras in an overlap-free area, and provides a human body target re-identification method among multiple cameras. The method comprises the following steps: A) dividing a known human body image sequence and an individual image sequence to be identified in a map depot into a plurality of groups, and calculating the individual similarity difference characteristics of each group according to prototype similarity characteristics; B) according to the individual similarity difference characteristics of each group and the corresponding tag of the individual similarity difference characteristics of each group, carrying out classifier training; and C) calculating the output weight average value of the classifier so as to determine the human body which is matched to a highest degree. The method can use multiple frames of image data to realize the human body re-identification and measure the similarity of a human body target in the multiple frames of image data. Compared with an algorithm provided by the prior art, a significant difference distance provided by the invention is higher in an identification rate on a data set, and the robustness of influence factors, including large size, dimension, shielding and the like, generated on the human body in a sample image among different cameras can be effectively improved so as to improve the identification rate of the algorithm.

Description

technical field [0001] The invention belongs to the field of image recognition, and in particular relates to a method and system for re-recognition of human body targets between multiple cameras without overlapping areas. Background technique [0002] At present, the existing human body re-identification methods almost use a query image to perform similarity matching with all the images in the gallery. However, these methods have inherent defects, that is, the appearance characteristics of the human body may change significantly between cameras. The principle of visual recognition points out that target recognition is a dynamic process, that is, it is necessary to look at two targets for a period of time to recognize whether they are the same target. In the prior art, a human body re-identification algorithm based on the image sequence is proposed by using the image sequence generated by the pedestrian tracking in each camera. The algorithm calculates the similarity between...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to provide a method and system for re-identification of human body targets between multiple sensors, aiming at solving the problem that existing technologies cannot realize human body re-identification with multi-frame image data, and measure the human body in multi-frame image data The problem of target similarity

Images