Volume I-3
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., I-3, 117-122, 2012
https://doi.org/10.5194/isprsannals-I-3-117-2012
© Author(s) 2012. This work is distributed under
the Creative Commons Attribution 3.0 License.
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., I-3, 117-122, 2012
https://doi.org/10.5194/isprsannals-I-3-117-2012
© Author(s) 2012. This work is distributed under
the Creative Commons Attribution 3.0 License.

  20 Jul 2012

20 Jul 2012

TRAJECTORY-BASED REGISTRATION OF 3D LIDAR POINT CLOUDS ACQUIRED WITH A MOBILE MAPPING SYSTEM

A. Gressin, B. Cannelle, C. Mallet, and J.-P. Papelard A. Gressin et al.
  • IGN, MATIS, 73 avenue de Paris, 94160 Saint-Mandé, France, Université Paris-Est

Keywords: point cloud, mobile mapping, registration, ICP, eigenvalues, dimensionality, neighborhood

Abstract. Thanks to a hybrid georeferencing unit coupling GNSS and IMU sensors, mobile mapping systems (MMS) with lidar sensors provide accurate 3D point clouds of the acquired areas, mainly urban cities. When dealing with several acquisitions of the same area with the same device, differences in the range of several tens of centimeters can be observed. Such degradation of the georeferencing accuracies are due to two main reasons: inertial drift and losses of GNSS signals in urban corridors. The purpose of this paper is therefore to correct these differences with an accurate ICP-based registration algorithm, and then to correct the MMS trajectory using these retrieved local transformation parameters.The trajectory loop information plays a key role for that purpose. We propose a four-step method starting from a 3D point cloud with overlapping parts, and the trajectory of the mobile mapping system. First, a polygonal approximation of the trajectory is computed in order to first divide the whole registration problem in local sub-issues. Secondly, we aim to find all the potential overlapping acquired areas between these segments using simple bounding box intersections. Thirdly, for each pair of overlapping areas, an efficient variant of the ICP algorithm is proposed to (1) prune cases where segments do not share point clouds of the same areas and (2) retrieve the transformation parameters, for real overlapping cases. Finally, all these transformations are linked together, and fed into a global distance compensation problem, allowing to adjust the MMS trajectories for several passages. As a conclusion, this method is successfully applied to data acquired over Paris (France) with the Stereopolis mobile mapping system.