期刊文献+

基于Kinect相机的苹果树三维点云配准 被引量:32

3D Point Cloud Registration for Apple Tree Based on Kinect Camera
下载PDF
导出
摘要 为建立具有真实彩色信息的果树三维点云形态结构模型,用Kinect相机获取不同视角下果树的原始三维点云,针对传统最近点迭代算法对待配准点云的空间位置要求苛刻的问题,提出了改进的点云配准算法。首先通过归一化对齐径向特征算法搜寻点云关键点,并使用快速点特征直方图描述子计算关键点处的特征向量。然后根据求得的特征向量估计2片点云关键点之间的空间映射关系,再基于随机抽样一致性算法提纯映射关系并完成点云的初始配准。最后利用最近点迭代算法完成点云的精确配准。实验结果表明,通过在最近点迭代算法前增加点云初始配准算法,有效地提高了点云配准的准确性和稳定性,能够对任意初始位置的2片点云进行准确匹配,平均配准误差为0.7 cm。 Aiming at establishing a 3D point cloud model of fruit tree with true color to provide scientific guidance for the production management of orchard,a research on the registration method for two pieces of 3D original point clouds of fruit tree obtained from different perspectives was carried out. The 3D raw point clouds of apple tree in two perspectives were obtained based on Kinect camera and information fusion technology. Firstly,the background removal and noise filtering approaches were used to implement a data pretreatment for each piece of raw point cloud,and every relative exact point cloud of single apple tree was acquired in each specific angle. Secondly,by using depth information of fruit tree's point cloud image and object boundary characteristics,the key points were extracted based on NARF( Normal aligned radial feature) algorithm. Meanwhile,the FPFH( Fast point feature histograms) descriptor was developed to obtain the characteristic vector for each key point. Thirdly,according to the characteristic vectors,the pairs of corresponding key points between two pieces of point cloud were estimated and extracted. And the spatial mapping relationships between two pieces of point cloud were calculated by validating and refining all pairs of corresponding key points based on the RANSAC( Random sample consensus) algorithm. Then the rotation matrix and translation vector between the two neighboring point clouds were computed,by which,the initial registration of two adjacent pieces of point cloud was achieved further. Finally,on the basis of the initial registration,two pieces of point cloud were fused into the same space coordinate system to complete their precise registration through applying the ICP( Iterative closest point) algorithm. This paper carried out the experiments based on the above algorithms,and the results showed that the improved point cloud registration method could be used to match two pieces of point cloud at any original positions in space,and its mean registration error reached 0. 7 cm.
出处 《农业机械学报》 EI CAS CSCD 北大核心 2016年第5期9-14,共6页 Transactions of the Chinese Society for Agricultural Machinery
基金 国家自然科学基金项目(31371532)
关键词 苹果树 Kinect相机 三维点云 初始配准 最近点迭代算法 apple tree Kinect camera 3D point cloud initial registration iterative closest point algorithm
  • 相关文献

参考文献21

二级参考文献221

共引文献151

同被引文献291

引证文献32

二级引证文献192

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部