Object-oriented residential building land-use mapping using lidar and aerial photographs

Human activities have transformed at least one-third of the Earth's surface in the past century, and land-cover and land-use analyses play a critical role in human-environment interaction analysis. Land-use analysis, especially for residential land uses, is comparably more challenging than land cover as land use categories relate less directly to the physical reflectance obtained by remote sensors; however, land-use classification recently attracts a growing attention because of the advancement in high-resolution imagery and the demand to improve intra-urban structure mapping. Current methods focus more on pixel-and parcel-based analysis but relatively less on object-oriented studies based on meaningful urban elements. This research presents an approach to detect and separate residential land uses on a building scale directly from remotely sensed imagery to enhance urban land-use analysis. Specifically, the proposed methodology applies a multi-directional ground filter to generate a bare ground surface from lidar data, then utilizes a morphology-based building detection algorithm to identify buildings from lidar and aerial photographs, and finally separates residential buildings using a supervised C4.5 decision tree analysis based on seven selected building land-use indicators. Successful execution of this study produces three independent methods, each corresponding to the steps of the methodology: lidar ground filtering, building detection, and building-based object-oriented land-use classification. Furthermore, this research provides a prototype as one of the few early explorations of building-based land-use analysis and a successful separation of more than 79.73% of residential buildings based on an experiment on an 8.25-km ² study site located in Austin, Texas.