Camera resource scheduling method and device in dynamic tracking

Abstract

The invention discloses a camera resource scheduling method and device in dynamic tracking, and the method comprises the steps: building video analysis tasks of N initial cameras so as to carry out the monitoring of a target, and determining a camera where the target is located from the N initial cameras; according to the obtained video analysis results of the N initial cameras, the target pixel coordinate position of the camera where the target is located is calculated, and the space coordinate position of the target is calculated according to the target pixel coordinate position; and querying a background database, finding out all associated cameras according to the spatial coordinate position of the target and a preset spatial association range, and scheduling the associated cameras tomonitor the target. According to the invention, reasonable and accurate task scheduling can be carried out on the camera by combining a background data management technology and a video positioning technology, and unnecessary video analysis is reduced, so that the cost of a dynamic tracking technology is greatly saved, and resources are ensured to be fully and effectively utilized.

Description

technical field [0001] The present invention relates to the technical field of video monitoring, in particular to a camera resource scheduling method and device in dynamic tracking. Background technique [0002] At present, machine vision has been widely used in the video surveillance industry, and object tracking is one of the typical applications. The deep learning algorithm is used to analyze the monitoring screen of the real-time camera, and the characteristics of the analysis result are compared with the characteristics of the tracking target. If the comparison result exceeds the preset threshold, the target is considered to be the tracking target. If all the cameras in the monitoring area are visually analyzed, and the spatial positions of the cameras are known, as long as you know which camera captured the target, you can know the current spatial position of the target, so as to achieve the purpose of target tracking. [0003] Visual analysis of all cameras in the mo...

AI Technical Summary

Problems solved by technology

[0004] However, the cost of the commonly used solutions in the prior art is relatively high. The face dynamic analysis server (analyzes the faces in the real-time video screen and performs target comparison) costs about 200,000 to 300,000 in 28 channels. In the case of the road, the cost of the face server alone can reach millions. If the human body structure analysis server is added, the cost is very expensive

In addition, in the actual tracking situation, if a camera that is far away from the target or a camera outside the field of view is used for analysis tasks, it will cause a waste of resources

Images