Multi-scale convolution feature fusion pedestrian re-identification method based on pose embedding

Abstract

The invention discloses a multi-scale convolution feature fusion pedestrian re-identification method based on pose embedding, and the method comprises the steps: carrying out the preprocessing of an original pedestrian image through a random erasing mode, obtaining a pedestrian image, carrying out the baseline network optimization of a Resnet-50 network model, and extracting deep convolution features; extracting a saliency human body image from the original pedestrian image; performing posture extraction on the human body saliency image, and then extracting local semantic features from the body part image; carrying out weighted fusion on the deep convolution features and the local semantic features, carrying out distance measurement on weighted fusion features, and generating an initial measurement list; and reordering the images in the initial measurement list according to a reordering algorithm to obtain an image correct matching ranking, and outputting a pedestrian matching image to identify a specific pedestrian. The identification and positioning precision can be greatly improved.

Description

technical field [0001] The invention belongs to the technical field of image processing methods, and relates to a multi-scale convolution feature fusion pedestrian re-identification method based on pose embedding. Background technique [0002] In recent years, artificial intelligence, as an important point of progress in the development of science and technology, has come out on top among all the technologies we use. Its use in the field of intelligent surveillance has also become extremely important. With the expansion of the city, the monitoring system is further popularized, and there are thousands of cameras all over the streets and alleys in every city. The use of cameras is increasing, and the cost of monitoring alone is extremely expensive, and it is impossible to monitor so many images at the same time. Therefore, pedestrian re-identification technology has attracted the attention of researchers. It can help people to monitor, track, and identify pedestrians. Hum...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a multi-scale convolution feature fusion pedestrian re-identification method based on pose embedding, which solves the problem of low accuracy of pedestrian re-identification caused by dislocation and background changes caused by pedestrian posture changes in the prior art question

Images