单光子激光雷达数据去噪与滤波算法

单光子激光雷达获取的点云数据存在大量噪声,这给数据的处理带来了挑战。基于局部距离统计提出了改进的点云去噪算法,对单光子激光雷达点云原始数据进行去噪。然后基于统计分析方法改进了点云滤波算法,对去噪后的点云数据进行滤波处理。利用新提出的算法与传统算法去噪和滤波后得到的点云进行比较,并与传统激光雷达的数字地形模型(DTM)数据进行对比。计算得到MABEL地面点云相对于传统激光雷达高程的均方根误差RMSE为2.98m,相关系数R^2为0.9938。进一步对地面点云插值得到剖面数字高程模型(DEM)数据,其相对于传统激光雷达高程的均方根误差RMSE为2.85m,相关系数R^2为0.9931。实验结果显示,提出的单光子激光雷达点云去噪和滤波算法优于传统算法,与传统激光雷达DTM数据具有较好的相关性,能够精确的恢复地形信息。

Research on De-noising and Filtering Algorithm of Single Photon Lidar Data

There is great amount of noise in the point cloud data acquired by single photon lidar, which brings challenges for data processing. Firstly, based on the local distance statistics, an improved point cloud de-noising algorithm is proposed to de-noise the original point cloud data of single photon lidar. Then, based on the statistical analysis method, an improved point cloud filtering algorithm is proposed to filter the de-noised point cloud data. The ground point cloud obtained after filtering algorithm is compared with digital terrain model(DTM) obtained from traditional lidar for quantitative evaluation of filter results. Calculation results show that the RMSE between MABEL ground point cloud and traditional lidar is 2.98m, and the correlation coefficient is 0.9938. Furthermore, linear digital elevation model(DEM) is obtained through interpolation of ground point cloud, the RMSE between linear DEM and traditional lidar is 2.85m and the correlation coefficient is 0.9931. Experiment results showed that single photon lidar point cloud processing results has a good correlation when compared with DTM data from traditional lidar. Thus, single photon lidar can provide accurate topographic information.