Auto-selection of airborne oblique images for automatic façade texture mapping

Geospatial applications based on photo-realistic building model are getting popular in the field of 3D GIS. Conventionally, the vertical imaging systems are designed for topographic mapping. Thus, it is improper for the extraction of good quality façade texture. On the other hand, for close-range photography, the occlusion and camera orientation problems will introduce not only texture quality issue but also cost-effective problem. The purpose of this paper aims on automatic selection of airborne oblique imagery for facade texture mapping provided that a large quantity of airborne oblique images with exterior and interior orientation parameters, lens distortion parameters and 3D polyhedral building models are existed. In this research, the airborne oblique images were taken from the helicopter with 30 degrees off-nadir angle and a flying height of 650 meters result in 13∼18cm ground sampling distance. Considering a large quantity of oblique images and 3D building models existed in a project, from the automation and efficiency point of view an efficient auto-selection algorithm is necessary. In the paper, many criteria were designed for the filtering of not visible images and for the selection of the best quality image without occlusion from other buildings. At first, for each façade, a visibility analysis is performed for all oblique images. It includes the visibility analysis of target façade within the imaging field-of-view and visible by the image. In order to avoid occlusion from other buildings, a filtering technique base on bounding sphere and bounding circle was developed. Finally, the average ground sampling distance and incident angle were considered for the selection of image with best quality. Experimental results depict that the proposed method can select the best qualified image without occlusion efficiently.