Pedestrian re-identification method, device and equipment and storage medium

Abstract

The invention discloses a pedestrian re-identification method, which comprises the steps of adding an instance normalization module into a neural network model to obtain an improved neural network model; training an initial feature extraction model according to a source data set and the improved neural network model; inputting a target data set into the initial feature extraction model to obtain pedestrian features; clustering the pedestrian features, and generating a label set corresponding to the pedestrian features according to a clustering result; adjusting the initial feature extraction model according to the target data set and the label set to obtain a trained feature extraction model; and recognizing pedestrians according to the trained feature extraction model. According to the pedestrian re-identification method disclosed by the invention, the generalization performance of the model and the performance in the target domain can be improved, and the identification accuracy andefficiency are improved.

Description

technical field [0001] The present invention relates to the technical field of image analysis, in particular to a pedestrian re-identification method, device, equipment and storage medium. Background technique [0002] The core goal of person re-identification is to determine whether there is a specific pedestrian in an image or video sequence. In recent years, with the emergence of large data sets and the development of deep convolutional neural networks, the performance of person re-identification has continued to rise. [0003] The current mainstream method is to construct a pedestrian re-identification dataset, and then use a deep neural network to train a feature extraction model on the dataset, and extract pedestrian features through the model. However, due to differences in camera resolution, camera angle, environmental background, lighting, season, etc. between different datasets, the distribution of data collected by different cameras in different datasets or even i...

AI Technical Summary

Problems solved by technology

However, due to differences in camera resolution, camera angle, environmental background, lighting, season, etc. between different datasets, the distribution of data collected by different cameras in different datasets or even in the same dataset is usually different.

Moreover, the existing models cannot handle changes in lighting and style well, which leads to a significant drop in pedestrian re-identification performance when the model trained on the source data set is directly applied to another target data set for pedestrian search. Decline

One solution is to manually label the target data set, and then use the target data set and artificial label information to fine-tune the existing model. However, labeling data is very time-consuming and expensive. In real applications, it is often necessary to arrange a large number of camera, the data distribution of each camera is inconsistent with the distribution of training data, and it is basically infeasible to label data for each camera

Images