Automatic Registration of Panoramic 2.5D Scans and Color Images

Abstract

Here we present an approach for a registration framework to fuse the 2.5D data from a laserscanner and the color information from a digital camera to 2.5D color images. The spatial resolution of the scanner is very dense, which emphasizes the need for precision during the calibration of both sensors as well as in the registration of the data streams. To achieve the same visual field for the digital camera, it is mounted on a vertical tilt unit. Overlapping color images are acquired by rotating the entire system horizontally and the tilt unit vertically by predefined angle increments. Using these angle increments as fixed input parameters, the calibration of the camera as well as the registration between both sensors can be simplified enormously. We use the Sift method and Ransac to find homologous points between the scanner and the camera data automatically and marker-free. In our results the marker-free approach is compared against marker-based registration.