| With the development of human society,people's living standards increased year by year,people's safety requirements are getting higher and higher.Indoor video surveillance industry developed in recent years,the annual indoor surveillance camera will shoot down the massive monitoring of video,but the huge video data is the video analysis and storage bottlenecks,a large number of unused video will be seriously consumed Storage of resources,it will also seriously affect the efficiency of the analysis.In this paper,a video segmentation system based on indoor video content is designed.The system can process the indoor surveillance video to find out the high-value video segment which is in existence and store it in a centralized manner.to reduce the storage pressure and improve the efficiency of the analysis.This paper shows a set of video cutting system,including algorithm design and system architecture design,is designed for the needs of real engineering and sample video.The algorithm uses multi-feature fusion method.include audio features and image features.The audio domain uses loudness and processing methods to reduce computational complexity.In the image domain,a video frame extraction method is proposed based on the motion consistency of the characters.Then,the algorithm of combining the Gaussian model with the HOG pedestrian detection is used to judge the movement of the characters,and the position and velocity of the characters are extracted.ANN to determine the video cut points.System architecture,the use of server clusters to build.The system consists of a business server,a database,a storage device,and an operating device.Through the analysis of the test video on the system,the statistical results and related data,it is concluded that the system can segment the video with high accuracy and can process the video in real time and reduce the consumption of the storage resources.The efficiency of artificial video analysis.Server system to meet the actual project needs. |
PDF Full Text Request