FACADE TEXTURE MAPPING FOR BUILDING MODELS USING TLS DATA AND CLOSE RANGE IMAGES

摘要

Facade texture mapping is an important task in the reconstruction of photo realistic building models. Close range images are always needed in texture mapping. However, building facades are frequently blocked by self-occlusion or foreground elements on images. Terrestrial laser scanning (TLS) data contains accurate 3D point clouds, which provide depth information to distinguish occlusions from building structures. Thus, the point clouds can be used in conjunction with multi-angle images to generate occlusion-free image patch for each building facade. The procedures are divided into three main steps. First, close range images and TLS data are registered in one coordinate system. Then, the occlusions are detected by the back-projection of foreground points. After occlusion removal, the image holes are compensated by counterpart images. The result indicates that combing TLS data and close range images is a feasible way to remove occlusions for mapping building facade texture.