Automatic Reconstruced Roof Shapes for Lidar Strip Adjustment and Quality Control

In the recent years, the performance of LiDAR systems was improved by acquisition of the terrain surface with steadily increasing point densities. However, the main error sources affecting the quality of LiDAR derived secondary products like DTMs and DSMs are resulting from systematic residual errors coming from insufficient calibration and strip adjustment, and from deficiencies in classification and filtering of the laser points. The systematic errors are recognized as discrepancies of the laser point clouds in overlapping areas of neighboring LiDAR strips. In this work, the main focus is on a new 3D measurement technique based on intersecting roof ridge lines and roof planes which are automatically reconstructed from the laser point clouds. The coordinate differences between conjugate intersection points are incorporated in an adjustment process to resolve for the residual errors of each LiDAR strip separately. The new 3D reconstruction method is applied on two different datasets, consisting of last pulse and full waveform data, and also taking advantage of full waveform measurements like intensity and pulse width which are decomposed from the waveforms. In general, the results show that significant discrepancies mainly in position still exist. After the strip adjustment and correction, the relative horizontal displacements between adjacent strips are improved significantly by more than 70%. The investigations also show that the higher laser point density of full waveform LiDAR data (3-5 points/m) leads to better results after the adjustment with respect to the last pulse dataset (density 1-2 points/m). * Corresponding author.