三维模型时空子空间引导的智能视频侦查系统

Smart video crime detection based on 3D model spatio-temporal subspace

为了克服传统视频处理技术面临的"语义鸿沟"等难题,借助三维模型时空子空间所蕴含的信息进行视频处理分析,提出了三维模型时空子空间引导的智能视频侦查技术。1在体形子空间的引导下从视频中匹配三维目标模型。2三维模型时空子空间引导下提取视频事件:监控对象视频+三维模型时空子空间→监控对象三维动作。3三维事件库中的动作比对分类:运动数据+三维事件库→视频类型和性质。文章涉及图形学、视频处理和刑事技术,探索了使用三维图形学技术解决视频侦查难题的新渠道。

In order to overcome the problem of the ＂semantic gap＂ faced by the traditional video processing technology, this paper proposes a three-dimensional model of spatio-temporal subspace guided smart video detecting technology. Its core idea is that the video data is processed and analyzed with the information contained in the 3D model of spatio-temporal subspace. This paper include： 1, matching 3D target model with the video under the guidance of the shape subspace; 2, 3D model of spatio-temporal subspace guided extraction of video events： monitor object video ＋ spatio-temporal subspace of 3D model → 3D monitored object movement; 3, Comparison of movements in 3D event Library： Sports data ＋ 3D event library → video type and nature. This paper is related to graphics, video processing and criminal technology. It establishes new channels for the use of 3D graphics technology to solve the problem of video detection and has an important academic significance.