基于多摄像头拼接的视频摘要  

Video abstract based on multi-camera video stitching

作　　者：宋戈[1] 胡伟[1] 肖碧波[1] 

机构地区：[1]北京化工大学,北京100029

出　　处：《电子设计工程》2014年第17期185-188,共4页Electronic Design Engineering

基　　金：国家自然科学基金(61003132)

摘　　要：根据sift经典算法和视频摘要原理,对多摄像头实时拼接在视频摘要上的应用进行了研究,通过将每组对应帧图像按照首帧图像配准后产生的变换矩阵和剪切模板进行变形及融合,使多摄像头实时合成一个;根据碰撞能量约束,对新摄像头中的运动物体在时间线上重新组合,完成视频摘要。实验利用了摄像头相对位置固定的条件,并充分考虑视频分辨率对实时效果的影响,与单摄像头相比,多摄像头动态拼接的视频摘要在单位时间内具有更高的信息量。According to the classical SIFY algorithm and the video abstract principle, the application of multi camera real-time mosaic in video summarization is researched. Each group of the frames captured from multiple cameras do deformation and cropping with the corresponding transformation matrices and the splicing template generated from the first frames matching, so the multiple cameras real-time becoming one camera is achieved; according to the collision energy constraint, the moving objects of the new camera are rearranged in the time line, then the video abstract is completed, the condition that the cameras are fixed in the relative positions is used, and the impact of video resolution for real-time results is given full consideration, Compared with single camera, the video abstract of the multi camera dynamic mosaic has more information in a unit of time.

关 键 词：多摄像头 实时 视频摘要 SIFT 

分 类 号：TN948[电子电信—信号与信息处理]