Roof segmentation using true-orthoimage and DSM with Obia paradigm

Currently analyzing high-spatial resolution image data to classify the object is one of the most important research topic. Conventional stereo-plotting method that use manual object delineation will take a long time in the process. Therefore, a more effective approach is needed to overcome it. OBIA (Object Based Image Analysis) is the solution offered to facilitate the identification of objects by considering to the shape, size, area, and information contained in the objects. In this study, we use true-orthoimage and DSM (Digital Surface Model), generated by UAV (Unmanned Aerial Vehicle) images, with 5 cm spatial resolution as the base data to get the segmentation results from the roof of the building. In which, DSM is used to obtain height information for each object. Moreover, DSM data is used to get the slope of the roof to separate different roof planes of the same building. Then, we apply segmentation process on the roof of the building based on the aspect of the roof. Segmentation of the object is done by multi-resolution segmentation technique by considering the parameters of scale, shape and compactness to get maximum accurate result in each building. This approach is applied to identify the roof of the building at the seashore villages of Tainan City, Taiwan with an area of 200 x 200 square-meters. Finally, the result of segmentation is compared with the manually delineation result for accuracy assessment.