Abstract:

As a leading method for capturing 3D urban scene data, laser scanning technology has been increasingly used in feature extraction, object recognition and modeling tasks. This study presents a new strategy that can be used to quickly and accurately extract building facade features based on point clouds, which are captured from laser scanners. The data first need to be pre-processed, including building a Kd-OcTree mixed index and calculating the normal vectors of point clouds using principal component analysis (PCA). On this basis, the initial clusters are obtained via fuzzy clustering, and then the generalized Hough transformation (GHT) is used in each cluster according to the sampling interval to detect the local peak value to obtain the preliminary plane. Next, similar planes are merged together based on the normal vectors and distance thresholds of the pending planes to generate better planeness. Finally, the extraction effects are optimized by the adjunctive judgment of neighborhood points, which is used to classify boundary points into the correct plane. The proposed approach has been tested with three different terrestrial laser scanners (TLS) datasets, and the results show that this mixed approach is able to speed up building facade extraction as well as the recall.