Detection of building changes from LIDAR data and aerial imagery

Change detection for land cover is an important work in the geo-information field. Traditionally, change detection is usually done by using multi-temporal images through the spectral analyses. Those images provide two-dimensional spectral information without including 3D shape in the third dimension. Among others, building models are the prominent object in the 3D world. As the availability and quality of emerging LIDAR (Light Detection and Ranging) systems that make the acquisition of shape information convenient, we include LIDAR data and aerial images to detect the changes of building models. The proposed scheme comprises four major parts: (1) data pre-processing, (2) detecting changes on old building areas, (3) treatment for unchanged buildings, and (4) finding new or changed buildings in the rest areas. The first step performs the spatial registration for the three types of data. In addition, we remove ground and vegetation areas. In the second step, we include LIDAR data and aerial images in the area of old buildings to detect changes. In the third step, if a building model is reckoned as unchanged, we mark the area in LIDAR to ensure finding new or changed buildings the next step. For the rest areas, we search for new or changed buildings by using mathematical morphology. The test site is located at Hsin-Chu city of north Taiwan. The old building models were acquired in 2002. The aerial images were acquired in 2005 with 12cm resolution. The LIDAR point clouds were also acquired in 2005 with a density of 1.5pts/m ². Preliminary results indicate that the result of detecting building changes by proposed scheme may reach high fidelity.