基于稠密光流的多媒体视频背景减除仿真

Multimedia Video Background Subtraction Simulation Based on Dense Optical Flow

传统方法对多媒体视频背景进行减除存在操作复杂、耗时久等问题,现提出基于稠密光流的多媒体视频背景减除方法。方法根据视频现场环境可编程门列阵的微小颗粒度并行特性,设计一种稠密光流算法,对视频全部像素对应的光流信息进行计算;采用ORB-SLAM技术创建全局视频环境的三维背景特征点云模型,将模型部分背景三维特征点云和全局视频环境的三维背景特征点相结合,引入到全局视频环境的背景信息中,运用三维均值漂移算法对视频中背景特征进行提取;最后对含有背景的区域与视频前景分离,并进行背景减除处理,由于减除后的视频含有较大的噪声干扰,因此还需对处理后的视频进行膨胀和腐蚀运算,最终得出平滑的多媒体视频。仿真发现,提出的视频背景减除方法鲁棒性较强、处理后的视频较为流畅,且整体运算量较小,效率较高。

Traditionally,the method of reducing multimedia video background leads to complex operation and high time consumption.This article puts forward a method to subtract multimedia video background based on dense optical stream.According to the micro-granularity parallel characteristic of FPGA in video field,this method designed a dense optical flow algorithm and calculated the optical flow information corresponding to all pixels of video.Then,the method used ORB-SLAM technology to create 3D background feature point cloud model of global video environment.Moreover,3D feature cloud of partial background was combined with 3D feature point of global video environment,which was introduced into the background information of global video environment.In addition,three-dimensional mean drift algorithm was used to extract the background feature of video.Finally,the region with the background was separated from the foreground of video,and then the background was subtracted.Because the reduced video included massive noise interference,it was necessary to expand and corrode the processed video.Finally,the smooth multimedia video was obtained.Simulation results verify that the robustness of proposed method is stronger,and the processed video is smoother.Meanwhile,the overall computation is smaller and the efficiency is higher.