第一人称动态延时摄影相机路径规划

First-Person Hyper-Lapse Video Camera Path Planning

第一人称动态延时摄影视频在加速播放时抖动剧烈,使人眼无法观看,传统路径规划技术在加速后仍然不能得到有效地优化加速视频。针对上述问题,提出一种基于滑动平均法与B样条曲线拟合的三维相机路径规划方法,利用相机三维空间坐标,将相机路径平滑问题转化为空间坐标平滑拟合问题,通过两步优化相机路径:首先通过滑动平均法计算近似平滑的路径坐标,与原坐标进行对比,筛选出偏差较小、靠近平滑路径的视频帧,然后通过B样条曲线对剩余的相机坐标进行拟合处理,以获取更为平滑的相机路径,确定输出的帧的位置。最后通过仿真,证明了上述方法的可行性。

The first-person hyperlapse video jitters violently during speed-up playback,which makes it impos-sible for human eyes to watch.Traditional camera path planning techniques can not optimize the accelerated first-person video effectively.To deal with this problem,we presented a camera path planning method based on moving average and B-spline curve fitting.The method transformed the problem of camera path smoothing into spatial dis-crete point fitting by using camera's space coordinates.We obtained approximate smooth path through moving average and deleted frames which deviate from approximate smooth path,then,obtained the smooth camera path by using B-spline curve fitting.This method can simplify the camera path smoothing problem,and it is easier to implement.