Surface Gradient Approach for Occlusion Detection Based on Triangulated Irregular Network for True Orthophoto Generation
MetadataShow full item record
Aerial images of urban areas have been used as base information for a diversity of applications. Considering the great quantity of tall buildings in these areas, it is important to have a method to automatically generate a product called true orthophoto mosaic, which represents all objects above the ground (buildings, bridges, etc.) in their true location. However, to create a true orthophoto, it is necessary to consider the occlusions caused by the surface height variation and to compensate for the lack of information using adjacent aerial images. The automatic occlusion detection is the bottleneck during the true orthophoto mosaic generation. The main aim of this paper is to introduce a new approach for occlusion detection - the surface-gradient-based method (SGBM) applied to a triangulated irregular network (TIN) representation. The originality of the SGBM is the occlusion detection principle, which is based on the concept of surface gradient behavior analysis over a TIN surface. The current methods interpolate a point cloud into a gridded digital surface model, which can introduce artifacts to the representation. The SGBM represents the surface as a TIN-based solid by taking into account the Delaunay constraint in the original point cloud, avoiding the interpolation step. The occlusions are then compensated using specific cost functions and refined via color blending. Experiments were performed and the results were assessed by using quality indicators (completeness), the consistency of orthoimage mosaic, and the time of processing. Experimental results demonstrated the feasibility of the SGBM for occlusion detection in the true orthophoto generation.