Object-based extraction of building features from LiDAR and aerial

The paper describes a newly developed object-based extraction method named MORPHSCALE. It is a multi-scale framework built based on mathematical morphology. First, LiDAR point clouds are classified to obtain Digital Terrain Model (DTM) and Digital Height Model (DHM). Second, the scale-space is generated for the DHM and the aerial photograph. Pulse intensity derived from LiDAR is used as the near-infrared channel and added into red, green, and blue channels of the aerial photograph. Third, the father-child relationship is built across the scale-space for the formed clusters. Based on this relationship, building features can be extracted. New multi-scale database structure is also introduced to store all components of the complex structures. MORPHSCALE is demonstrated using LiDAR and multi-spectral aerial photograph but it promisingly works well with any DHM and spectral sources. Future works will assess the relationship between the extracted spectral property of each building feature and its roof material