【摘要】 针对光子计数激光雷达数据特点，研究基于泊松分布的点云去噪算法并开展精度评估。首先，将点云投影到二维剖面，划分格网并统计每个格网光子点个数，剔除点数大于平均值部分以计算背景噪声率；随后，从小到大调整格网尺寸，统计各尺寸下格网内的点数，大于阈值时将该网格内的点都标记为信号，并且根据比例大小划分为高、中、低置信度三类；最后，采用分段直线拟合将倾斜点投影到直线上以识别倾斜地形，采用分段二次拟合方法剔除残余孤立噪点，得到优化结果。利用多组不同地形光子点云数据开展实验，结果表明：基于泊松分布的去噪算法在冰盖、海洋场景下效果较好，整体精度优于96%,在植被场景稍差，但能达到识别信号的基本目标。更多还原

【Abstract】 According to the characteristics of photon counting LiDAR data, the point cloud denoising algorithm based on Poisson distribution is studied and the accuracy is evaluated.Firstly, the point cloud is projected onto the two-dimensional section, the grid is divided, and the number of photon points in each grid is counted.The part whose number of points is greater than the average value is eliminated to calculate the background noise rate; then, the grid size is adjusted from small to large, and the number of points in the grid is counted at each size.All points in the grid are marked as signals when they are larger than the threshold and classified into three categories of high, medium and low confidence according to the scale; finally, the inclined points are projected onto the line by piecewise linear fitting to identify the inclined terrains.The piecewise quadratic fitting method is used to eliminate the residual isolated noise, and the optimization results are obtained.Experiments are carried out using several groups of photon point cloud data with different terrain.The results show that the Possion algorithm works better in the ice caps and oceans scenes, with an overall accuracy better than 96%,and slightly worse in the mountains, but it can also achieve the basic goal of signal recognition.更多还原