Video pedestrian re-identification method based on twin double-flow 3D convolutional neural network

Abstract

The invention relates to a video pedestrian re-identification method based on a twin double-flow 3D convolutional neural network. The video pedestrian re-identification method based on the twin double-flow 3D convolutional neural network comprises the steps of extracting each frame of picture of a pedestrian video into an optical flow-x feature map, an optical flow-y feature map, a gray feature map, a horizontal coordinate gradient feature map and a vertical coordinate gradient feature map through a hard line layer of the twin double-flow 3D convolutional neural network; taking the optical flow feature map as input of an action branch to extract action information of the pedestrian, and taking other feature maps as input of an appearance branch to extract appearance information of the pedestrian; fusing the pedestrian action information into the extracted pedestrian appearance information; performing metric comparison learning on the action information and the appearance information through fusion; updating the network parameters, and training a new convolutional neural network; and associating the target pedestrian image with the to-be-identified pedestrian image with the first similarity ranking. Compared with the prior art, the method has the advantages of being closer to a real scene and the like.

Description

technical field [0001] The invention relates to the field of machine vision based on image processing, in particular to a video pedestrian re-identification method based on a twin-stream 3D convolutional neural network. Background technique [0002] Person re-identification, a problem faced in person matching on non-overlapping cameras, has received increasing attention in recent years due to its importance in implementing automated surveillance systems. The re-identification of video pedestrians is closer to the real scene. The invention is helpful to realize the intelligentization of the city, to the safety and tracing of large public places such as airports, and to the automatic search for lost old people and children through the camera. To assist the public security organs in the automatic identification and tracking of criminals. [0003] In many applications, such as cross-camera tracking and pedestrian search, it is desirable to recognize a person from a group of peo...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a video pedestrian re-identification method based on twin-stream 3D convolutional neural network in order to overcome the defect of low recognition degree in the busy environment of the above-mentioned prior art

Images