【摘要】 针对点云配准算法易受噪声、体外孤点以及采样率影响的问题，采用形状指数关键点检测方法、最近邻距离比法和迭代最近点算法，基于三维点云对羊体点云配准方法进行研究。结果表明：1)使用协方差矩阵特征描述子能对形状指数方法检测的关键点进行描述；2)基于特征匹配的配准方法能对不同视角的羊体点云进行配准，最大均方根误差为0.024 1;3)对于含有噪声、体外孤点或较低采样率的不同类型的羊体点云模型，配准的最大均方根误差为0.023 8。试验证明基于特征匹配的配准方法能准确地对羊体点云进行配准，并且不受噪声、体外孤点以及采样率的影响。更多还原

【Abstract】 Aiming at the problem that the point cloud registration algorithm is easily affected by noise, outliers and sampling rate, the shape index key point detection method, nearest neighbor distance ratio method and iterative nearest point algorithm are adopted to study the point cloud registration method of sheep based on three-dimensional point cloud. The results show that: 1) The feature descriptor of covariance matrix can describe the key points detected by shape index method; 2) The registration algorithm based on feature matching can register point clouds of sheep at different view angles, and the maximum root mean square error is 0.024 1; 3) The maximum root mean square error of registration is 0.023 8 for different types of sheep point cloud models with noise, outliers or low sampling rate. The experiment proves that the registration method based on feature matching can accurately register sheep point cloud, and is not affected by noise, outliers and sampling rate.更多还原