针对传统机载LiDAR点云滤波算法中，因种子点选取存在随机性，导致地面点出现错分类与漏分类的问题，该文提出一种融合高精度三维模型与数字正射影像(DOM)的点云滤波方法。该方法首先基于高精度三维模型和DOM数据，构建搜索起始基准面并进行地形区域划分；然后建立全区坡度图；最后根据坡度信息确定最优分类种子点，实现地面点精确提取。以大崇镇消落带数据进行实验和分析，结果表明，该方法可有效避免种子点选取随机的问题，减少点云地面点错分类和漏分类现象，提升了点云地面点自动提取的准确度和作业效率，为后续高精度地形建模与工程应用提供了可靠的数据基础。

To address the issues of random selection of seed points in existing point cloud filtering algorithms, as well as wrong classification and missed classification of ground points, this paper proposed a point cloud filtering method that integrated high-precision 3D models and digital orthophoto map(DOM). Firstly, based on high-precision 3D models and DOM data, a reference plane for searching is constructed and areas are divided. Secondly, a slope map for the entire area is created. Finally, based on the slope information, seed points of optimal classification are determined to exactly extract ground points. The results show that the method could effectively reduce random classification of seed points, wrong classification and missed classification of ground points in point cloud based on experiments and analysis conducted using data from the water-level-fluctuation zone in Dachong Town, improving the accuracy and operational efficiency of automatic ground point extraction and providing a reliable data basis for high-precision terrain modeling and engineering application in the future.