Automated texture mapping of 3D city models with images of wide-angle and light small combined digital camera system for UAV

Abstract

To overcome the disadvantages of single non-metric camera with small format and short baseline, a wide-angle camera system is developed with the combination of four digital cameras (UAV-LAC) to meet low altitude photogrmmetric for acquiring both facades and rooftops high resolution texture of 3D city model. A relative self-calibration method based on tie points in the overlapping areas of each two sub images is used to make compensation for the deformed errors due to the light and simple constructed mechanical frame. AAT is used to rapidly retrieve pose parameters of mosaic image and original inclined images. After that, space forward intersection algorithm is used to gradually improve the building model as well as matching each space edge of building. Next, geometrical rectification is considered for textures and the best textures for wall and roof are selected by taking into account occlusion, image resolution, surface normal orientation, and coherence with neighboring triangles. Finally, by incorporating friendly necessary human interaction into texture reconstruction algorithm, an semiautomatic system is designed to reconstruct texture from UAV-LAC.