摘要
针对激光雷达点云数据缺乏纹理信息的问题,该文提出一种基于互信息的车载激光雷达点云与全景影像配准方法。该方法使用统一的球面全景成像模型,引入互信息作为相似性测度,将车载激光雷达点云生成的深度图与信息提取后的全景影像进行配准,实现配准参数的自动、高精度解算。同时,对车载激光雷达点云与全景影像配准的精度进行评定与分析。实验结果表明,车载点云与全景影像的配准方案是可行的,具有较高的配准精度。
The registration of vehicle-borne laser point clouds and panoramic image is critical during the data processing of vehicle-borne mobile mapping system. A method based on mutual information for the registration of panoramic image and vehicle borne laser point clouds was presented. The method used uni- fled spherical projection as the imaging model and mutual information as the similarity measurement with the registration of depth maps derived from vehicle-borne laser point clouds and segmented panoramic ima- ges, getting automatic and high-precision registration coefficient. Meanwhile, the evaluation of registra- tion accuracy was analyzed. As the experimental results suggested, the registration method presented by this paper was feasible and comparatively high-precision, which provided a wide application in urban 3D modeling and reconstruction.
出处
《测绘科学》
CSCD
北大核心
2016年第4期113-117,123,共6页
Science of Surveying and Mapping
基金
国家科技支撑计划项目(2012BAJ23B03)
关键词
车载激光雷达
点云
全景影像
配准
互信息
vehicle-borne I.iI)AR
point cloud
panoramic image
registration
mutual information