A new generation of space-borne LiDAR (Light Detection And Ranging) satellite ICESat-2 (Ice, Cloud, and land Elevation Satellite-2) equipped with ATLAS (Advanced Topographic Laser Altimeter System) can perform earth observation. The main problem is to remove the noise photons from the data. The study proposes a main direction-based noise removal algorithm based on three sets of photon-counting LiDAR data. In order to extract the main direction, features in the spatial neighborhood (k) of photons are calculated, most of the initial noise is removed according to the angle between the main direction of photons and the along-track distance direction. Qualitative and quantitative evaluations are employed to validate the proposed algorithm. The obtained results and the performed analysis reveal that the proposed algorithm can process day and night data with different signal-to-noise ratios, while the accuracy of various surface types exceeds 96%. More specifically, the accuracy of the proposed algorithm for night data can reach 97.43%. Based on quantitative evaluations using SPL (Single photon LiDAR), MATLAS, and airborne LiDAR data, the average R, P, and F values are 0.951, 0.959, and 0.954, respectively. Meanwhile, the result of the proposed algorithm is compatible with the ATL03 photons with low, medium, and high confidence, and its accuracy is superior to ATL08 products. The proposed algorithm had fewer parameters and significantly outperformed the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and the improved local statistical distance algorithm. This algorithm is expected to provide a reference for subsequent photon-counting LiDAR data processing.

新一代星载LiDAR（光探测和测距）卫星ICESat-2（冰、云和陆地高程卫星2号）配备ATLAS（先进地形激光高度计系统），可以进行对地观测。主要问题是从数据中去除噪声光子。该研究提出了一种基于三组光子计数激光雷达数据的基于主方向的噪声去除算法。为了提取主方向，计算光子的空间邻域（ k ）中的特征，根据光子主方向与沿轨距离方向之间的角度去除大部分初始噪声。采用定性和定量评估来验证所提出的算法。所得结果和分析表明，该算法可以处理不同信噪比的昼夜数据，且各种地表类型的准确率超过96%。更具体地说，所提出的算法对于夜间数据的准确率可以达到97.43%。根据SPL（单光子激光雷达）、MATLAS和机载激光雷达数据进行定量评估，平均R 、 P和F值分别为0.951、0.959和0.954。同时，该算法的结果与低、中、高置信度的ATL03光子兼容，精度优于ATL08产品。该算法参数较少，明显优于基于密度的噪声应用空间聚类（DBSCAN）和改进的局部统计距离算法。该算法有望为后续光子计数激光雷达数据处理提供参考。