Method for reliably generating video abstraction in complex scene

Abstract

The invention discloses a method for reliably generating video abstraction in a complex scene. The method comprises the following steps: performing background modeling on an acquired video to obtain the moving foreground of a current frame in a video sequence; selecting a proper target detector for performing target classification and detection on the moving foreground of the current frame by using prediction information; calculating the movement locus of a moving target in the current frame by using a multi-target tracking method according to the acquired type and detection result of the moving target, and calculating the prediction information of a next frame; detecting a next frame till the end of the video sequence; generating video abstraction according to information such as the movement loci, types and the like of all moving targets. According to the method, a mutual feedback mechanism is introduced into foreground detection and tracking processes, so that the omission ratio of weak targets, the false drop rate of dense shielded targets, and the fault tracking probability, miss tracking probability and incomplete probability of the loci of moving target can be reduced effectively; the method can be applied to various complex scenes.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for generating reliable video summaries in complex scenes. Background technique [0002] In modern society, video surveillance systems are playing an increasingly important role in all walks of life. Departments such as public security, fire protection, transportation, and industrial production have particularly urgent needs for video surveillance in public places. Video surveillance systems have become an important part of maintaining social order. , strengthening social management and an important part of safe production. The number of surveillance cameras is increasing rapidly, and a large amount of surveillance video data is generated every day, which leads to a large amount of manpower, material resources and time spent on finding clues through surveillance videos, and reduces the effective utilization of surveillance videos. According to the statistics of...

AI Technical Summary

Problems solved by technology

[0005] Video summarization technology is of great significance and has broad market prospects for the effective use of massive surveillance videos. However, the existing video summarization technology cannot effectively solve the problems of weak and small targets in complex scenes, and the mutual occlusion and adhesion of targets. In order to quickly browse the surveillance video, Reduce the missed detection rate and false detection rate of moving target event extraction in complex scenes, reduce the probability of tracking errors and missing targets in the process of trajectory generation of moving targets, and eliminate users' concerns about existing video summarization technology.

Images