The Registration of Laser Scanner Three-dimensional Point Clouds Using Two-dimensional Images Recorded With a Sensor Combination Camera

The laser scanning method has been used for three-dimensional (3-D) modelling in many applications for over two decades. Laser scanners collect spatial point data from the scan area in a short time, and laser scanning data have their own local co-ordinates in each station. The other hand, 3-D modelling applications require many scans from different stations to obtain occlusion free 3-D model. In such a case, all scans must be registered into a common reference co-ordinate system. Registration of point cloud into the reference co-ordinate system is still an important research topic in photogrammetry and computer sciences fields. Many methods have been developed for registration of point clouds. Nevertheless, if there is no implicit entity on the scan area, these methods are not practicable because corresponding points can not be selected from the point clouds. In this study point clouds were registered by the image recorded by the sensor combination camera. Although many laser scanners have integrated camera, a high resolution digital camera is usually mounted to the laser scanner for efficient 3-D modelling. The image based registrations with the sensor combination of laser scanner and camera were performed by making use of a single image, relative orientation and bundle adjustment triangulation methods. The results were compared with the iterative closest point (ICP) which is the most widely used method for registration of point clouds.