大数据视频图像关键帧的检索模型优化仿真  

Big Data Video Image Retrieval Model Optimization Simulation of Key Frames

作　　者：张波[1] 王颖[1] 

机构地区：[1]郑州大学西亚斯国际学院,河南郑州451150

出　　处：《计算机仿真》2016年第9期446-449,共4页Computer Simulation

基　　金：基金项目:基于ArcGIS的数字化城市系统中某些关键技术的研究与实现(152102210132)

摘　　要：传统的检索方法在建模时,主要通过视频图像特征提取视频图像的关键帧,忽略了视频图像采集过程中受到光照等因素的影响,使得图像特征发生变化情况,导致建模误差大的问题。提出基于视频元数据信息的大数据视频图像关键帧的检索建模方法。先对采集的样例大数据视频进行图像划分,获取图像子区域的灰度均值,将编号相同的子区域直接进行连接,组成大数据图像视频的关键帧时序特征曲线,通过异常因子去除视频图像中的突变干扰,在引入视频元数据信息算法计算视频图像关键帧元数据,并进行视频元数据信息对比,建立大数据视频图像关键帧的检索模型。仿真结果表明,采用改进算法检索效果好,误差低。In this paper, we proposed a retrieval model of key frame in big data video image based on the video metadata information. Firstly, the image partition of big data videos of collected sample was conducted to obtained the grey average value of image subdomain. Then we connected the subdomain with the same number directly to build time-order feature crave of key frame in big data video image and wiped the mutation interference in video image using abnormal factor. Moreover, we introduced the algorithm of video metadata information to calculate the key frame metadata and compared the video metadata information. Finally, we built the retrieval model of key frame in big data video image. The simulation results show that the modified method has better retrieval effect and small error.

关 键 词：时序特征 子片段 大数据视频 

分 类 号：TN911.73[电子电信—通信与信息系统]