A waveform exploitation optimized multiscale curvature algorithm for recognizing ground echoes from airborne laser scanner in densely forested area

Extracting reliable digital terrain models is a significant prerequisite for utilizing the airborne laser scanning data in various applications. The most widely used ground filtering algorithms only consider the spatial and structural information from geometry perspective of LiDAR data and perform poorly in densely forested area where the laser shot cannot penetrate heavy canopy to detect the true ground. A waveform exploitation optimized multiscale curvature algorithm is presented to introduce contributions from radiometry of LiDAR in recognizing vegetation signals to complement spatial metrics based filtering approaches. Preliminary results of this novel strategy have shown its potential in classifying the point cloud data in dense vegetation area, even when the low vegetation clusters are distributed very close to the true ground and misleading for spatial information based algorithms. It reduces both the commission and omission errors significantly in areas where the true ground beneath the dense canopy top is under sampled.