This study introduces a method for filtering lidar data based on Fuzzy C-Mean (FCM) clustering. This method is composed of four key steps. In the first step, a Digital Surface Model (DSM) is generated for the first and last pulses. In the second step, the generated DSM and the lidar intensity image are reshaped and applied as input data for a FCM clustering process to automatically classify buildings, trees, roads and ground. In the third step, The DSM pixels which correspond to roads and ground in the classified image are interpolated into a grid DTM while the pixels which correspond to buildings and trees are omitted from the interpolation process. Finally, the interpolated DTM is smoothed by a low-pass filter to remove low vegetation and other objects which might be classified as ground. Datasets from mountainous and flat urbanized areas were selected to test the proposed filter. To meet the objectives, the generated DTM was compared against reference data that was generated manually and both omission and commission errors were calculated. Experimental results suggest that, compared with the widely applied progressive TIN (triangular irregular network) densification (PTD), the FCM approach is able to reduce omission, commission and total errors by 7.06%, 7.17% and 5.26% respectively in the case of urbanized area, and by 4.25%, 2.02% and 1.81% respectiveley in the case of mountainous areas. |
