Pedestrian re-recognition method based on unsupervised deep model and hierarchical attributes

Abstract

The invention discloses a pedestrian re-recognition method based on an unsupervised deep model and hierarchical attributes. The method includes the steps of first, pre-processing a pedestrian picture and dividing the pedestrian picture into a plurality of overlapping blocks according to body parts; second, constructing a convolutional neural network (CNN) with three hidden layers, training the model through a stacked convolutional auto-encode (CAE), and pre-training and fine-tuning the model using CUHK and VIPeR pedestrian data sets, respectively; third, designing hierarchical attributes for the pedestrian and a classifier for each attribute, and inputting pedestrian features extracted from the CNN to each attribute classifier to obtain the probability of the corresponding attribute; and fourth, obtaining the posterior probability of the category by combining an attribute category mapping relation to determine the category to which the sample belongs. According to the invention, the problem of lacking a labeled training sample is effectively solved by pre-training the CNN model with an unsupervised learning method of the CAE; the accuracy of pedestrian re-recognition is effectively improved by using the characteristics that the CAE can be used to well reconstruct an image; and the re-recognition of pedestrians is more in line with the law of human cognition by introducing the hierarchical attributes, so that the pedestrian re-recognition method is allowed to have a semantic expression ability and a practical application value.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a pedestrian re-identification method based on an unsupervised depth model and hierarchical attributes. 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 crowded areas. Places prone to public safety incidents. The emergence of surveillance cameras has undoubtedly brought great convenience to people. Surveillance video can provide public security departments with clues to major criminal cases such as shopping mall theft, gang fights, and bank card theft; at the same time, it can also provide a large amount of real-time traffic information for traffic coordination departments to facilitate ...

AI Technical Summary

Problems solved by technology

Therefore, only using the attributes of a single layer for re-identification greatly limits the application range of person re-identification.

Images