Target detection method, image recognition server and system

Abstract

The invention discloses a target detection method, an image recognition server and a system, wherein the method comprises the following steps: S1, receiving a monitoring image, wherein the monitoringimage is divided into a RoI region and a non-RoI region, the RoI region is compressed according to a first compression rate, the non-RoI region is compressed according to a second compression rate, and the first compression rate is less than the second compression rate; S2, performing image recognition for the monitoring image through a preset image recognition model, and according to a result ofimage recognition, sending a compression rate updating parameter for adjusting the first compression rate and/or the second compression rate to a sending end of the monitoring image. The method, the server and the system, provided by the invention, have the advantages of greatly reducing data transmission quantity, having low requirement on networks, greatly reducing network cost, improving efficiency of image detection, responding quickly, having high detection precision, and so on.

Description

technical field [0001] The invention relates to the field of image target detection, in particular to a target detection method, an image recognition server and a system, and is especially suitable for target detection of a distributed monitoring system of the Internet of Things. Background technique [0002] Intelligent video surveillance technology is a relatively new research field developed on the basis of computer vision and image processing technology, combined with other related technologies and theories. It aims to use the data analysis capabilities of computers or intelligent processing units to automatically realize static or The perception, description and analysis of dynamic physical objects meet the needs of daily production, intelligent security in life, intelligent transportation, and urban intelligent construction. [0003] The image target detection system distinguishes between centralized and distributed in terms of computing modes: [0004] 1) Centralized...

AI Technical Summary

Problems solved by technology

Surat Teerapittayanon of Harvard University and others proposed "Distributed Deep Neural Networks over the Cloud, the Edge and End Devices" at the IEEE 37th International Conference on Distributed Computing Systems in 2017, which uses a distributed system architecture, but it can only be used for a single object. Image classification, since target detection requires classification and positioning of objects in the image, the above system cannot be regarded as a distributed image target detection system in the strict sense

Images