改进动态半径的矿井激光雷达点云滤波算法  

Point cloud filtering algorithm of mine LiDAR using improved dynamic radius

作　　者：景宁波[1] 马宪民[1] 郭卫[2] 秦学斌[1] JING Ningbo;MA Xianmin;GUO Wei;QIN Xuebin(College of Electrical and Control Engineering,Xi'an University of Science and Technology,Xi'an 710054,China;College of Mechanical and Engineering,Xi'an University of Science and Technology,Xi'an 710054,China)

机构地区：[1]西安科技大学电气与控制工程学院,陕西西安710054 [2]西安科技大学机械工程学院,陕西西安710054

出　　处：《西安科技大学学报》2023年第2期406-413,共8页Journal of Xi’an University of Science and Technology

基　　金：国家自然科学基金项目(51905416)。

摘　　要：由于煤矿井下粉尘和水雾环境的影响,激光雷达采集的点云数据存在多源混合噪声。针对动态半径滤波方法在处理激光点云数据时存在近场噪声过多和远场有效数据丢失的问题,提出一种改进动态半径的矿井激光雷达点云滤波算法。结果表明:采用正交线性变换主成分分析法对三维点云数据进行降维处理,得到多组二维数据;对每组二维数据进行拉普拉斯运算,识别高频噪点,对数据进行第1次滤波;使用基于扇环邻域的统计分析,删除数据中的异常点,对数据进行第2次滤波,将经过2次滤波后的多组二维数据恢复成三维点云数据。在矿井尘雾噪声环境中,该算法获得98.9%的滤波准确率,计算时间稳定在0.5 s以内。对比统计滤波、半径滤波和动态半径滤波算法,在提高激光点云滤波准确率和改善点云结构化特征的同时,该算法减少运行时间,有效抑制点云中的尘雾噪声。Due to the influence of dust and fog environment in coal mine,there existsmulti-source mixed noise in the point cloud data collected by LiDAR.In order to solve the problem of too much noise in the near field and loss of effective data in the far field when dynamic radius filtering is used to process point cloud data,a improved dynamic radius filtering algorithm based on fan ring neighborhood is proposed.Firstly,the orthogonal linear transformation principal component analysis method was used to reduce the dimension of 3D point cloud data with multiple sets of 2D data obtained;Secondly,Laplace operation was performed on each group of 2D data to identify high-frequency noise in the data with the data filtered for the first time;Then,using the statistical analysis based on the fan ring neighborhood,the outliers in the data were deleted with the data filtered for the second time.Finally,multiple sets of 2D data after twice filtering were restored to 3D point cloud data.The filtering accuracy of the algorithm was 98.9%,and the calculation time was stable within 0.5 s under the experimental environment of mine dust and fog noise.The results indicate that compared with statistical filtering,radius filtering and dynamic radius filtering algorithms,the propased algorithm can improve the accuracy of point cloud filtering and enhance the structural characteristics of the point cloud,shorten the running time,and effectively suppress the dust and fog noise in the point cloud.

关 键 词：激光雷达 尘雾噪声 拉普拉斯算子 动态半径滤波 统计滤波 

分 类 号：TD76[矿业工程—矿井通风与安全] TP391.41[自动化与计算机技术—计算机应用技术]