Hierarchical clustering of 3-D line segments for building detection

A novel approach for an efficient extraction of rectangular boundaries from aerial image data is proposed in this paper. In this approach, a Centroid Neural Network (CNN) with a metric of line segments is utilized for connecting low-level linear structures or grouping similar objects. The proposed an approach, called hierarchical clustering method, utilizes the fact that rooftops of a building are about the same height and perform clustering process with candidate 3-D line segments with similar heights. Experiments are performed with a set of high resolution satellite image data. The results show that the proposed hierarchical clustering method can remove noisy segments such as shade lines efficiently and find more accurate rectangular boundaries.