摘要
三维点云的相关处理可以应用于机器人目标识别、裂纹检测等方面。几乎每一项实际应用都需要对点云进行特征提取,目前已经出现了许多种特征提取方法。为了研究对比各种点云特征提取方法的性能,基于PCL点云库,设计算法对三种常用的特征提取描述子的描述性、紧凑性、鲁棒性、时间效率等性能指标进行了评估,并给出了绘制精度召回率曲线的实现方法。实现了基于不同特征描述子的点云配准,并计算配准结果误差。通过综合性能评估和配准结果对比,分析了各种特征提取算法的应用场合以及局限性。
The relevant processing of the 3D point clouds can be applied to target recognition and crack detection.As almost every practical application requires feature extraction of point clouds,many feature extraction methods have been proposed,and each method has its own applicable field.To study and compare the performance of different feature extraction methods,different descriptors are computed and algorithms are designed to assess the performances of descriptors in this paper.The performance indicators including descriptiveness,compactness,robustness against Gaussian noise and time efficiency are evaluated based on the PCL point cloud library.Method of drawing precision recall curve is presented.Point cloud registration is performed based on different descriptors and the result errors are computed.The application scenarios and limitations of the descriptors are analyzed through the comprehensive performance evaluation and registration results comparison.
作者
单光胜
赵岚
Shan Guangsheng
出处
《工业控制计算机》
2021年第10期74-76,79,共4页
Industrial Control Computer
