Comparison of the ground filtering and dtm generation algorithms on uav-based 3D point clouds

In this study, 1D Progressive Morphological (PM1D), 2D Progressive Morphological (PM2D), Maximum Local Slope (MLS), Elevation Threshold with Expand Window (ETEW) and Adaptive TIN (ATIN) ground filtering algorithms were used to filter the UAV-based point clouds with four different densities (raw, low-density, medium-density and high-density data). Performances of these algorithms were investigated in terms of the Type I, Type II, Total Error, Kappa Index and Chi-squared values. The DTM (Digital Terrain Model) of the study area was then generated by interpolating the most accurate filtering result in MATLAB environment by using 16 interpolation algorithms namely orthogonal (linear, quadratic, cubic) polynomial interpolation algorithms, non-ortogonal (bi-linear, bi-quadratic, bi-cubic) polynomial interpolation algorithms, Shepard, Radial-basis functions (multiquadratic-circular right cone, multiquadratic-circular paraboloid, multiquadratic-circular hyperboloid of two sheets, inverse multiquadratic, gauss, cubic spline, natural cubic spline, multilog) and ordinary kriging (OK) algorithms. Vertical accuracy of the produced DTM was examined with terrestrial measurements. It was also investigated that whether or not UAV-based point clouds can be used to generate DTMs as an alternative for LiDAR (Light Detection and Ranging) point clouds.