A method and device for pedestrian re-identification based on clustering algorithm

Abstract

The present invention disclosed a method and device based on a cluster -based algorithm. The re -identification method of the pedestrian includes: S1: extraction of pedestrian characteristics of every pedestrian in different scenarios; S2: According to the number of scenes and each oneThe number of pedestrians in the scene establishes a two -dimensional matrix; S3: According to the two -dimensional matrix established, the two -dimensional distance matrix of the characteristics of pedestrian characteristics in different scenarios; S4: The density cluster of the two -dimensional distance matrix, the output distance is less than the preset threshold value than the preset thresholdThe feature collection; each feature collection of the output is the pedestrian characteristics of a pedestrian in different scenarios; S5: The output feature set is marked in the corresponding scenario to obtain the position of the same party in different scenarios.The present invention can re -recognize the characteristics of the same party in many scenarios, and can identify the characteristics of repeated pedestrians through the method of clustering without specifying the number of pedestrians.

Description

Technical field [0001] The present invention relates to the field of computer technology, and more particularly to a pedestrian reinacity method and apparatus based on a clustering algorithm. Background technique [0002] With the popularity of artificial intelligence technology and depth learning framework, pedestrian reconscribed increasingly challenging topics in computer visual fields. [0003] In the conventional technique, it can be generally divided into two methods, one is a method based on the colored regional characteristics, and cut out the target rectangle from the video pattern as an input image, extracted with the foreground extraction and online clustering. In the color area, the statistical characteristics of the color area are used as partial features to the human re-identification; the other is the secondary optimization method of the monitoring video pedestrian re-identification result of the time and space constraint, using the human appearance visual characte...

AI Technical Summary

Problems solved by technology

[0004] However, the method of pedestrian re-identification based on color region features uses the color of the pedestrian's own clothes as the main comparison object, which also leads to the problem that the re-identification of pedestrians with similar colors in multiple scenes is not obvious enough. , hairstyle, whether to wear glasses, whether to carry a bag, etc., will lack consideration. In addition, the problem of changing the color of clothing for the same pedestrian on the way cannot be solved; and the secondary optimization method for the re-identification results of surveillance video pedestrians constrained by time and space, The image input process is required to first mark the time, and the pictures with scattered time nodes will not be recognized. Secondly, the space is also required. It is necessary to know the pedestrian path in advance, and then sort it in time order

Images