Roof Detection in Lidar Data

Lidar is widely used in many fields in recent years. Consequently, research of feature extraction in lidar data has intensified. Roof of building as a stable line feature is widely used in many fields. But there is yet not a good algorithm to finish this work. In this paper, we propose a new method to extract roof of building. The roof is modeled by a symmetric exponential roof edge model and the altitude image which generated from original lidar point cloud data is smoothed by a low-pass filter ISEF which is optimal for the symmetric exponential model. And then an algorithm for roof detection and a grouping and fitting method are proposed for line feature extraction. In order to depress the effect of the noise a fusion method is used for multi-images. In the end of the paper the method is proved useful through the lidar data comes from Calgary University in the end of the paper.