改进影像匹配点云滤波方法的下沉盆地建模

The modeling of subsidence trough based on an improved image matching point cloud filtering method

针对受地表植被物候性特征的影响,无人机影像匹配点云所生成DEM精度偏低的问题,该文提出了一种基于坡度阈值和相邻点高差阈值相结合的地面种子点选取方法,改进了移动曲面拟合点云滤波算法。以林南仓煤矿西四采区为例,基于无人机测绘技术获取实验区地表三维点云,结合实测数据,分别采用目视检验、定量评价和DEM精度分析等验证了滤波算法的性能。结果表明,改进算法的滤波总误差平均值为13.43%,在大部分地形条件下滤波效果良好,可有效剔除地表植被点云,提高实验区DEM精度,DEM中误差为±5.9 cm,与真实地形的相关度达到0.998 7。通过改进算法建立的下沉盆地模型能够满足矿区地表变形监测精度要求,可快速求取下沉盆地形态参数。

Due to the phenological characteristics of surface vegetation,the accuracy of DEM generated by UAV image matching point cloud is low.A ground seed point selection method based on the combination of slope threshold and adjacent point height difference threshold was proposed in this paper,and the filtering algorithm of moving surface fitting point cloud was improved.Taking the west No.4 mining area of Linnancang Coal Mine as an example,the three-dimensional point cloud of the surface was acquired based on UAV mapping technology,combined with the measured data;the performance of the filtering algorithm was verified by visual inspection,quantitative evaluation and DEM accuracy analysis.The results showed that the average filtering error of the improved algorithm was 13.43%,and the filtering effect was good under most terrain conditions.The surface vegetation point cloud could be effectively eliminated and the DEM accuracy of the experimental area was improved.The error in the established DEM was ±5.9 cm and the correlation of the real terrain reached 0.998 7.The subsidence trough model established by the improved algorithm could meet the accuracy requirements of surface deformation monitoring in the mining area,and he morphological parameters of the subsidence trough could be quickly obtained.