| 三维点云的分割及骨架提取研究 |
| |
| 引用本文： | 范保青,姚剑敏,林志贤,严群,李成跃.三维点云的分割及骨架提取研究[J].测绘科学,2021,46(1):162-169,195. |
| |
| 作者姓名： | 范保青 姚剑敏 林志贤 严群 李成跃 |
| |
| 作者单位： | 福州大学物理与信息工程学院,福州350116;福州大学物理与信息工程学院,福州350116;福州大学物理与信息工程学院,福州350116;福州大学物理与信息工程学院,福州350116;福州大学物理与信息工程学院,福州350116 |
| |
| 基金项目： | 福建省科技重大专项;广东省光信息材料与技术重点实验室开放基金资助项目;广东省科技重大专项;国家重点研发计划课题项目 |
| |
| 摘 要： | 针对在三维点云环境下分离目标物体所出现的过度分割问题,提出一种结合随机抽样一致性和颜色差值区域聚类的分割方法。首先利用RANSAC算法去除场景中大部分平面,使得目标物体和连成片的点云脱离,然后结合点云的距离阈值和目标颜色差值,得到目标点云数据。针对L1中值算法对曲率较大模型的骨架提取存在的不足,进行了改进。通过L1中值算法对点云模型进行骨架提取,得到点云的骨架点,然后沿端点方向向外进行最大内切球的球心提取,最后连接多个球心及骨架末端点,得到符合人类视觉效果的骨架。改进的算法提高了L1中值对曲率较大点云骨架提取的准确性。
|
| 关 键 词： | Kinect深度图像 三维点云 点云分割 改进骨架提取 |
Research on segmentation and skeleton extraction of 3D point cloud |
| |
| FAN Baoqing,YAO Jianmin,LIN Zhixian,YAN Qun,LI Chengyue.Research on segmentation and skeleton extraction of 3D point cloud[J].Science of Surveying and Mapping,2021,46(1):162-169,195. |
| |
| Authors: | FAN Baoqing YAO Jianmin LIN Zhixian YAN Qun LI Chengyue |
| |
| Institution: | (College of Physics and Information Engineering,Fuzhou University,Fuzhou 350116,China) |
| |
| Abstract: | Aiming at the excessive segmentation problem that occurs when separating target objects in a three-dimensional point cloud environment.A segmentation method combining RANSAC and Color Difference Region Clustering Algorithm is proposed.Firstly,the RANSAC algorithm is used to remove most of the planes in the scene,so that the target object and the connected point cloud are separated,then the distance threshold and the target color difference of the point cloud are combined to obtain the target point cloud data.The deficiencies of L1median algorithm for the larger curvature model skeleton extraction are improved.The L1 median algorithm is used to extract skeleton of the point cloud model,and obtain point cloud skeleton points,then the center of the largest inscribed sphere is extracted outward along the end direction,Finally,this method connects the center of the ball and the end points of skeleton to obtain a skeleton that conforms to human visual effects.The improved algorithm improves the accuracy of L1-median extraction of large curvature point cloud skeleton. |
| |
| Keywords: | Kinect depth image 3Dpoint cloud point cloud segmentation improved skeleton extraction |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
