REGISTRATION OF TIME OF FLIGHT TERRESTRIAL LASER SCANNER DATA FOR THE STOP-AND-GO MODE H. M. Badawy *, N. M. Alsubaie, M. Elhabiby**, N. El-Sheimy *Department of Geometrics Engineering, University of Calgary, 2500 University Dr. NW, Calgary, AB T2N 1N4, Canada (hmmohamm, nalsubai, mmelhabi, elsheimy)@ ucalgary.ca ** Public Works Department, Ain Shams University, Cairo, Egypt Commission VI, WG VI/4 KEY WORDS: Stop-and-Go, Time of Flight, Registration, Terrestrial Laser Scanner, Mobile Laser Scanning, Time of Flight. ABSTRACT: Terrestrial Laser Scanners (TLS) are utilized through different data acquisition techniques such as Mobile Laser Scanning (MLS) and the output can be used in different applications such as 3D city modelling, cultural heritage documentations, oil and Gas as built, etc... In this research paper, we will investigate one of the modes of TLS on mobile mapping platform. Namely the Stop-and-Go (SAG) mode. Unlike the continuous mode, the Stop-and-Go mode does not require the use of IMU to estimate the TLS attitude and thus in- turn it has an overall reduction in the system cost. Moreover, it decreases the time required for data processing in comparison with the continuous mode. For successful use of SAG mobile mapping in urban areas, it is preferred to use a long range time of flight laser scanner to cover long distances in each scan and minimize the registration error. The problem arise with Long range laser scanners is their low point cloud density. The low point cloud density affects the registration accuracy specially in monitoring applications. The point spacing between points is one of the issues facing the registration especially when the matching points are chosen manually. Since most of TLS nowadays are equipped with camera on-board we can utilize the camera to get an initial estimate of the registration parameters based on image matching. After having an initial approximation of the registration parameters we feed those parameters to the Iterative Closest Point algorithm to obtain more accurate registration result. 1. INTRODUCTION TLS nowadays are vital for mobile mapping applications. One of the most interesting applications is 3-D modelling of large areas. To cover a large area usually full mobile mapping with a moving car with Inertial Measurement Units (IMUs) and Global Navigation Satellite Systems (GNSS) on-board and a short range Phase Shift (PS) laser scanner. The choice of the PS laser scanner for full mobile mapping is obvious to cope with the speed of the moving vehicle. That comes with the price of an expensive system and shortage in coverage especially with high buildings. To overcome the drawback of the full mobile mapping, the long range Time of Flight (ToF) laser scanner is used with the SAG mode. SAG mode requires registration of the point clouds obtained from each scan (stop). The registration error increases with the increase of number of scans. In the case of ToF long range TLS, a few number of scans can cover a large area that requires many scans if we use PS laser scanners. The problem arising with ToF data is their low density. This low density lowers the accuracy of the registration especially when matching points that are manually selected. Camera nowadays are associated with many TLS and obtains images at each scan. These images can be utilized to obtain the relative orientation of the TLS at each scan. In this research paper we investigate the automatic registration of ToF laser scanner point cloud. We perform automatic registration based on the image matching for images associated with each scan. The most efficient algorithm used nowadays for image matching is the Speeded-Up Robust Features (SURF) algorithm. After using the SURF algorithm we use the tie points in each * Corresponding author. image to estimate the relative orientation between images. Relative orientation parameters are then fed to the Iterative Closest Point ICP algorithm to find the best registration parameters for the scans. 2. AUTOMATIC REGESTRATION OF POINT CLOUD The registration process usually involves manual selection of matching points between the scans to be registered (at least three non collinear points to apply Least Squares (LS)). Manual selection of the matching points is an inaccurate and time consuming process, especially when the number of points is limited. The technique used to overcome the inaccuracy of the manual registration is to use a transformation parameters obtained from the manual registration and fed it to the ICP algorithm. Nowadays most of the Laser scanners are equipped with metric cameras. The images associated with each scan can be utilized to obtain their relative orientation parameters. These parameters could be used as an estimate for the initial transformation parameters used in the ICP algorithm. At least five tie points must exist between the stereo images in order to obtain the relative orientation. In this paper, the tie points are determined via the feature selection algorithm known as SURF (Bay, Ess, Tuytelaars, & Van Gool, 2007) . In the following subsection we are going to review the algorithms used to obtain perform automatic registration of point cloud using images. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-1, 2014 ISPRS Technical Commission I Symposium, 17 – 20 November 2014, Denver, Colorado, USA This contribution has been peer-reviewed. doi:10.5194/isprsarchives-XL-1-287-2014 287