Pedestrian re-recognition method based on pixels and channel attention mechanism

Abstract

The invention discloses a pedestrian re-recognition method based on pixels and a channel attention mechanism. The pedestrian re-recognition method comprises the steps of extracting global features ofpedestrians according to a boundingbox (retrieval box) of a person; averagely dividing the pedestrian picture into two parts and three parts, and respectively extracting local features of pedestrians;and matching the extracted character features with the character information in the Galertness, and finding out the required character information. A channel and a pixel attention module are used forfeature extraction, so that the influence of background information on a retrieval result is effectively reduced; meanwhile, middle-layer supervision is further designed for the neural network, in the feature extraction process, a multi-loss function is used for supervising middle-layer feature information, and network convergence is accelerated; according to the pedestrian re-recognition networkbased on the channel attention mechanism, the pixel attention mechanism and the middle layer supervision, redundant information in figure boundbox can be effectively deleted, figure information is effectively aggregated, and therefore retrieval precision is remarkably improved.

Description

technical field [0001] The invention relates to a pedestrian re-identification method, in particular to a pedestrian re-identification method based on a pixel and channel attention mechanism, and belongs to the technical field of image processing. Background technique [0002] At present, various crimes at home and abroad pose a great threat to the sustainable and stable development of society. In shopping malls, stations, airports, pedestrian streets and other places with a relatively large flow of people, there are large and small monitoring devices all over the place, but how to accurately find out the people or information we need from these monitoring information is still a huge challenge. Especially in the criminal investigation work, the police need to find the criminal suspect information from a large amount of long-term monitoring information, understand the situation in time and control it. However, the amount of monitoring information is huge, the content is comp...

AI Technical Summary

Problems solved by technology

Especially in the boundingbox (retrieval box), which contains character information and background information, the neural network cannot distinguish between the two, resulting in the background information being also part of the character features during feature extraction, which will affect the accuracy of the entire re-identification system have a huge negative impact

How to effectively reduce the impact of background information on re-identification technology is a huge challenge

Images