A method of person re-identification based on multi-component self-attention mechanism

Abstract

The invention provides a pedestrian re-identification method based on a multi-component self-attention mechanism. The method first pre-trains the backbone model of a deep convolutional neural network; Self-attention features; then input the multi-component self-attention features into the classifier, and jointly train to minimize the cross-entropy loss and measurement loss; finally, input the test set pictures into the trained model, and fuse the output component features to obtain the overall features, which are measured Sorting implements person re-identification. By fully considering various challenges in the pedestrian re-identification problem, the present invention proposes a multi-component self-attention mechanism, which effectively expands the attention activation area and enriches the characteristics of pedestrians; the self-attention module enables the network to pay more attention to The area with discriminative characteristics is where the spatial attention module and channel attention module are integrated into the network in the form of residuals, making the network more robust and stable, and easy to train.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and computer vision, and in particular relates to a pedestrian re-identification method based on a multi-component self-attention mechanism. Background technique [0002] With the acceleration of urbanization, public safety has become the focus and demand of people's increasing attention. Many important public health areas such as university campuses, theme parks, hospitals, and streets are widely covered with surveillance cameras, creating good objective conditions for automated surveillance using computer vision technology. [0003] In recent years, person re-identification, as an important research direction in the field of video surveillance, has attracted increasing attention. Specifically, pedestrian re-identification refers to the technology of using computer vision technology to judge whether a specific pedestrian exists in an image or video sequence under cross-camera and...

AI Technical Summary

Problems solved by technology

[0004] In an open environment, due to the complex and changeable monitoring scene, the collected pedestrian images often have interference factors such as background noise, illumination changes, posture changes, and severe occlusions. The area of ​​the degree of discrimination leads to the extraction of features that are not robust enough and the recognition performance is poor

Images