EFFICIENT BUILDING CATEGORY CLASSIFICATION WITH FAÇADE INFORMATION FROM OBLIQUE AERIAL IMAGES

摘要: Abstract. Building category refereed to categorizing structures based on their usage is useful for urban design and management can provide indexes of population, resource environment related problems. Currently, the statistics are mainly collected by manual from street data or roughly extracted remote sensing which either laborious too coarse. With (e.g. satellite aerial images), buildings be automatically identified top-view, but detailed categories single not recognized. Facade oblique-view image greatly help us identify buildings, example, balcony usually exist in resident buildings. Hence, this paper, we propose an efficient way classify building with facade information. Firstly, following texture mapping procedure, each building’s textures cropped oblique images via a perspective transformation. Then, average colour, stander deviation R, G, B channel, rectangle Haar-like features feed further random forest classifier identifications. In experiment, manually selected 262 facades that classified into four functional types as: 1) regular residence ; 2) educational building; 3) office 4) condominium. The results shows that, 30% as training samples, classification accuracy reach 0.6 promising real applications believe more sophisticated feature descriptors classifiers, e.g., neuronal networks, much higher.