针对现有地面点云简化算法存在点云边缘特征易丢失、特征点位识别不准确和点位空间分布不均匀等问题，本文提出了一种兼顾地形特征和点位均衡的LiDAR地面点云简化方法。该方法首先加权耦合多种地形因子，生成全面描述地形复杂变化的综合因子；然后根据各点边缘特征距离及特征显著性选取初始特征点集；最后构建综合地形特征约束的回归预测模型，迭代预测以捕捉准确、均匀的地形特征点。选取8组具有不同地形特征的地面点云数据作为研究对象，并将本文方法与7种代表性方法对比。结果表明：本文方法精度最高，生成的数字高程模型平均均方根误差和平均绝对误差分别降低了2.7% ~ 61.2%和2.0% ~ 61.9%，派生品（平均坡度和地形粗糙度）与真值也更为接近。本文方法对地面点云的简化结果最优，但在效率方面与先进算法相比还有待提升。

(Objective) The highly redundant ground point cloud data cannot ensure the efficiency of data storage, transmission and processing. Therefore, how to achieve intelligent and automated simplification of high-redundancy and high-complexity ground point cloud data has become a pressing problem to be solved in the field of surveying and remote sensing. (Method) In view of the problems existing in the current ground point cloud simplification algorithms, such as the easy loss of point cloud edge features, inaccurate identification of feature points, and uneven spatial distribution of feature points, this paper proposes a LiDAR ground point cloud simplification method that takes into account both terrain features and point distribution uniformity. This method first builds a comprehensive terrain feature factor based on the existing terrain feature factors, through the coupling of terrain feature selection and feature weighting, to construct a multi-dimensional description of the complex terrain features in the ground point cloud, avoiding the deficiency of a single terrain feature factor in describing various types of terrain features; then, the constructed comprehensive terrain feature factor is combined with the regression prediction model to build a regression prediction model considering terrain features, and at the same time, this paper defines the edge feature distance and feature significance based on the local geometric features of the point cloud, selects the initial feature point set to prevent the occurrence of edge contraction phenomenon, and takes this point set as the initial training data for the regression prediction model; finally, an iterative selection strategy for terrain feature points is proposed, using the regression prediction model to iteratively predict the values of the comprehensive terrain feature factors of each point in the ground point cloud, and obtaining the point with the largest prediction error as the terrain feature point, until the specified number of simplified points is reached. (Result) This paper selects eight sets of high-density LiDAR point cloud data from various different acquisition platforms, each presenting different local terrain features.The method proposed in this paper is systematically compared with seven representative existing methods (Uniform, MIWSA, TKPCS, FPUC, AdaC2F-TF, MQ-G, and IC2F) under different point cloud retention rates (1.0% to 0.1%) to verify the calculation accuracy and efficiency of the proposed method. The experimental results show that under different point cloud retention rates, the accuracy of the proposed method is the highest. The root mean square error (RMSE) and mean absolute error (MAE) of the generated digital elevation model are reduced by 2.7% to 61.2% and 2.0% to 61.9% respectively, and the derived products (average slope and terrain roughness) are also closer to the true values. The simplification result of the ground point cloud by this method is the best, but in terms of efficiency, it still needs to be improved compared with advanced algorithms. (Conclusion) In conclusion, the LiDAR ground point cloud simplification method proposed in this paper, which takes into account both terrain features and point distribution uniformity, not only effectively retains key terrain features, but also demonstrates high modeling accuracy and terrain expression ability under different point cloud densities, providing reliable technical support for high-precision digital terrain modeling.