Digital terrain model extraction from airborne LiDAR data in complex mining area

Airborne light detection and ranging (LiDAR) proved to be an adequate technique to deliver highly accurate 3D mass points of the surface. However, the surface of mining area is complex with steep slope, dense vegetation, artificial mining facilities and buildings, which is different from the flat surface of city. The main processing workflow for DTM generation from LiDAR includes points filtering and DEM interpolation. In this article, five methods for removing object points from LiDAR data in mining area were compared. These methods, including Adaptive TIN (ATIN), Elevation Threshold with Expand Window (ETEW), Maximum Local Slope (MLS), Mathematical morphology (MM), Iterative Polynomial Fitting (IPF), analyze data points based on variations of local slope, elevation and height difference between points and the interpolated surfaces. Complex mining area data set with various cliffs of quarry, trees, houses, roads and small reliefs were selected to test the filtering methods. The results show that all methods can effectively remove most object points in complex mining areas. The ATIN and MM filter generated the best result in sharp cliff area of a quarry, whereas the other algorithms tended to remove the steep edge of quarry and roads. Depending on the filtering parameters, each method experienced various omission or commission errors. Quantitative assessment shows the ATIN and IPF based on the height difference between points and surface perform better. DEM interpolation assessment experiments indicate that interpolation biases were minute. Global statistics show that Modified shepard's method, Spline and Radial basis function interpolation methods have the lowest errors in the study area.