The main challenge to generate virtual reality from the satellite and aerial imagery is the difficulty of reconstruction of 3D models in complex urban areas. one promising solution to the problem is the insertion of the 3D models of cultural features in the image data textured onto terrain elevation values. However, the necessity of extracting features manually in this method results in its failure to meet the automation requirement for the feature extrusion application. NETECH proposes a novel pattern recognition technique to automatically extract the features, therefore, the insertion of 3D models to the elevation data becomes feasible. Combining with a fast wavelet correlation method for stereo correspondence, the proposed approach can furnish the automatic feature extrusion with realistic appearances. The multiscale techniques and the parallel data structures of the underlying algorithms provide the advantages of utilizing multiple processors, which enables the system to significantly increase the throughput. During the Phase I effort, NETECH will prove the feasibility of the approach by prototype demonstration, which includes image registration, feature modeling, pattern recognition, stereo correspondence, and 3D scene rendering.
Keywords: Virtual Reality Object 3d Modeling Image Registration Terrain Visualization
