Distributed twin convolutional neural network pedestrian re-identification method based on cloud end, edge end and equipment end

Abstract

The invention relates to the technical field of electronic communication, in particular to a distributed twin convolutional neural network pedestrian re-identification method based on a cloud end, anedge end and an equipment end. According to the method, a distributed structure is utilized, so that all input data are not uploaded to the cloud for processing when the ReID problem is solved, and the ReID problem is solved at the local end and the edge end as much as possible. Specifically, exit points are arranged at three ends; joint training is carried out to obtain an excellent neural network model meeting the requirements of the invention. According to the method provided by the invention, the ReID recognition precision is improved, the data communication cost is greatly improved, the method can be properly improved and expanded to a multi-region camera network, the application of the ReID in reality is realized in a distributed manner, and particularly, the method has a wide prospect in the aspects of urban security and protection and crime fighting.

Description

technical field [0001] The present invention relates to three distributed terminals of cloud, edge and device terminal, twin convolutional neural network (Siamese Convolutional Neural Network, referred to as SCNN), pedestrian re-identification (Re-Identification, referred to as ReID) field, specifically a A distributed Siamese convolutional neural network pedestrian re-identification method based on cloud, edge and device. Background technique [0002] In recent years, with the rise of neural networks, the use of deep learning methods to solve ReID problems has become more and more accepted by many experts and scholars, and has greatly improved the recognition accuracy of ReID to a certain extent. At the same time, with the rise of the Internet of Things, camera deployment is more common, which provides the possibility to solve the ReID problem on the device side. However, there are still many problems in solving ReID on the device side and in the cloud: [0003] 1) If the...

AI Technical Summary

Problems solved by technology

[0003] 1) If the device-side data is offloaded to the cloud for processing, it will bring large communication costs, delays and privacy issues;

[0004] 2) If ReID is solved nearby on the device side, the accuracy of the ReID results may not meet the requirements due to the limitation of the number of network layers caused by the memory on the device side.

Images