基于三维点云的物体6D位姿估计方法

6D Object Pose Estimation Method Based on 3D Point Cloud

物体6D位姿估计近年来在计算机视觉及其他领域有着广泛的应用。为了掌握目前基于三维点云的物体6D位姿估计方法研究现状和趋势,对传统和基于深度学习的物体6D位姿估计的有代表性的方法进行汇总和分析。从基于三维模板、基于对应关系、基于投票三个种类的方法入手,详细介绍物体6D位姿估计方法的实现方式和优缺点,同时对其他最新的位姿估计方法作简要介绍。通过对基于三维点云输入的物体6D位姿估计方法的使用场景的总结,剖析了现存的问题,也提出了对未来发展方向的展望。

6D object pose estimation has been widely used in computer vision and other fields in recent years.In order to grasp the current research status and trend of 3D point cloud-based 6D object pose estimation methods,the traditional and representative methods of 6D object pose estimation based on deep learning are summarized and analyzed.Starting with three kinds of methods based on 3D template,correspondence and voting,the realization methods,advantages and disadvantages of the 6D object pose estimation method are introduced in detail,and other latest pose estimation methods are briefly introduced.By summarizing the application scenarios of the method for estimating the 6D object pose based on the input of 3D point clouds,the existing problems are analyzed,and the future development direction is also put forward.