Co-registration of photogrammetric and laser scanner data for generation of 3D visual models

Abstract

Nowadays, most terrestrial laser scanner (TLS) systems provide the facility of mounting a digital camera on the laser scanner. This not only facilitates a means to generate better quality photorealistically textured 3D models from TLS point clouds; it also offers the opportunity of using photogrammetric orientation techniques to complement existing methods of TLS point cloud registration. This paper describes an approach whereby a registration procedure based upon photogrammetric means is employed as the first step in integrating TLS data and imagery for the generation of textured 3D models. The approach, called image-based registration (IBR), entails an estimation of transformation parameters between the individual scan data and between digital imagery using photogrammetric bundle adjustment. Once both TLS and photogrammetric data are registered in the same coordinate system, the process of forming a segmented structured surface model and its associated triangular mesh are carried out. Photogrammetrically derived constraints are used to convert the unstructured, registered laser scanner model to a structured model. Finally, texture mapping takes place via the rectification of image patches from the integrated images used in the IBR process onto individual surface elements. Test results obtained with the proposed approach are presented to highlight its practicability and accuracy.