X. Jin and W. Sun (PRC)
Texture mapping, 3D model, texture images, laser synchronized visual images, error estimation
How to map texture images into 3D model is one of the most important problems in 3D real visualization of large scale outdoor scenes. However, the existing texture mapping methods are limited in accuracy by the estimation errors of camera parameters and the imaging difference between the laser data and the texture images. In this paper, a new method based on the platform of TH 3DLCS-2001, a 3D laser-visual synchronized scanning system developed by Tsinghua University, is proposed. This method introduces a low resolution visual image which is synchronously scanned with the laser data as intermedia between the laser data and high resolution texture images, to estimate and compensate the errors of texture mapping. The new method adopts the following steps. Firstly, the laser data and the laser-synchronized visual image are registered, to get the coordinate relationship between them. Secondly, the camera parameters are estimated using Tsai’s method. Then, utilizing the laser-synchronized visual image as intermedia, the global and local mapping residual errors between high resolution texture images and the laser data are calculated. Finally, after error compensation, the high resolution texture images are mapped into the 3D model. Experimental results show that the accuracy of texture mapping can be improved greatly.
Important Links:
Go Back
