Training method, recognition method, device and equipment of pedestrian re-recognition model

Abstract

The invention discloses a training method, a recognition method, a device and equipment for a pedestrian re-recognition model, and the method comprises the steps: obtaining a training data set, inputting the training data set into an enhanced visual perception learning network model, and obtaining a first recognition result; the feature extraction module performs feature extraction on the training data set to generate a first feature map; the attention module carries out fusion processing on the first feature map to obtain a second feature map, and spatial attention information and channel attention information are utilized; the batch attention module performs fusion enhancement processing on the second feature map to obtain a third feature map; the enhanced visual perception module carries out feature fusion processing on the second feature map and the third feature map, and a fusion function is enhanced through cross-feature connection; the recognition module recognizes the fourth feature map to obtain a first recognition result, and corrects the enhanced visual perception learning network model according to the first recognition result and the loss function to obtain a pedestrian re-recognition model, and the pedestrian re-recognition method can be widely applied to the field of image processing.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a pedestrian re-identification model training method, recognition method, device and equipment. Background technique [0002] Pedestrian re-identification (Person Re-identification, RE-ID) is a challenging task in computer vision, its goal is to identify the same target person by matching pedestrian images captured by non-overlapping surveillance cameras. Surveillance applications play a prominent role, including multi-camera activity analysis, subway video surveillance systems, multi-camera tracking and crowd counting. With the vigorous development of my country's smart city construction, the requirements for public safety are also increasing. In some public areas such as railway stations, roads, and airports, there are large-scale video surveillance network coverage. On the one hand, diverse and complex The environment poses challenges because lighting, resolution, person poses ...

AI Technical Summary

Problems solved by technology

With the vigorous development of my country's smart city construction, the requirements for public safety are also increasing. In some public areas such as railway stations, roads, and airports, there are large-scale video surveillance network coverage. On the one hand, diverse and complex The environment poses challenges because lighting, resolution, person poses and light conditions can vary widely among unseparated cameras. On the other hand, the clothes of different pedestrians are highly similar, which also creates a Considerable difficulty, so pedestrian re-identification faces technical difficulties

The existing CNN-based traditional method of visual perception is applied to pedestrian re-identification. The processing of overall features and local features is homogeneous, and insufficient attention is paid to salient areas. The perception ability is poor, and it is limited to mining rough information. At the same time, existing deep learning methods also fail to enhance the discrimination and richness of attentional knowledge in a clear way.

Images