使用卷帘快门相机进行移动测量时,由于逐行曝光的特点,影像上会产生果冻效应.而传统的运动恢复结构(Structure from Motion,SfM)算法假设影像是全局快门获取,直接处理卷帘快门影像很难得到高精度的结果.针对以上问题,构建了卷帘快门SfM框架,主要包括对初始化重建、卷帘快门影像绝对位姿估计、卷帘快门三角化和光束法平差等关键步骤的算法改进,分析并验证了卷帘快门SfM的退化问题,提出利用多镜头组合相机采集的影像进行卷帘快门SfM可以避免退化,并且利用多镜头约束的光束法平差对SfM结果进行进一步优化.实验表明,所提出的面向多镜头组合相机的卷帘快门SfM算法能够有效应对果冻效应,提高重建精度.
Rolling Shutter SfM Algorithm for Multi-Lens Combination Cameras
Images might suffer from rolling shutter effect due to the characteristic of progressive exposure if a rolling shutter camera for movement measurement is used.Traditional SfM algorithm typically assumes that images are obtained by a global shutter,so it is diffi-cult to obtain high-precision results when using them to process the rolling shutter images.To solve this problem,a rolling shutter SfM framework is constructed,which mainly includes algorithm improvements for key steps such as initialization reconstruction,rolling shutter image absolute pose estimation,rolling shutter triangulation and rolling shutter bundle adjustment.Then,the degradation of rolling shutter SfM is analyzed and verified.Images captured by multi-lens combination cameras are used to perform rolling shutter SfM to avoid degradation,and multi-lens constrained bundle adjustment is used to further optimize the SfM results.Experiments show that the rolling shutter SfM algorithm for multi-lens combination cameras proposed can effectively deal with the rolling shutter effect and im-prove the reconstruction accuracy.
photogrammetry and remote sensingrolling shutter SfMmulti-lens combination camerasbundle adjustmentrolling shutter
万方数据
维普
