Research On Key Techniques Of Data Processing From Mobile Mapping System For Land Vehicle

This dissertation presents a study of state-of-the-art Mobile Mapping System for Land Vehicle (MMS LV). The key techniques of data processings include strategies of real-time GPS-Aided Inertial Navigation Integration for POSLV, post-processing methods of POS (GPS/IMU) data, sensor calibration methods of lever arm and boresight misaligned angles, point matching based on sequence images in multiple views, calculating of three-dimensional coordinates based on Space Intersection, poles detection and road marking extraction from point clouds of LiDAR (Light Detection And Ranging). Based on the upper major aspects, a set of experiments and analysis have been made. Some useful conclusions which adapt to project application are educed. The followings are what we have done in this dissertation:1. The major works focus on the space relationship of position and orientation among Digital Camera, Laser,GPS,DMI and IMU sensors, and then calibration methods of diverse sensor's lever vector(lever arm and mounted angles) are discussed. A new calibration method is proposed to registration images with LiDAR's point clouds which are partly filtered or classified, and multi-sensors integration spatial registration is refined. For the requirements of project tests, calibration sites of MMS and distribution scheme of ground control points are carefully designed.2. According to data acquisition from two experimented projects, POS data post-processed techniques and methods, such as SingleBase, Differential GNSS, Precise Point Positioning (PPP) and so on, are investigated. Real-time navigation integration strategies of POSLV equipped with GPS/IMU/GAMS/DMI sensor systems are analyzed. Moreover, during GPS outages or without GPS outages,position and orientation accuracy about POS data post-processing is analyzed and compared.3. By applying the photogrammetric theory of space intersection to sequence images obtained by land vehicle, calculation formula is educed to get the correct 3D coordinates of spatial point. Two space intersection algorithms, where one is Space Intersection of Two Camera and One Location (SITCOL) and another is Space Intersection of One Camera and Two Location (SIOCTL), are implemented.4. The computer vision theory is introduced into homogeneous image points matching based on of sequence images in multi-views. Feature points detection by Harris operator, correlation coefficient matching and multiple views matching based on RANSAC solidity estimation are carried out through tests and analysis. Corresponding image points can be matched by two-view fundamental matrix (F), trifocal tensor (T) over three views, camera matrix and correlative theory or method in image sequence. 5. Considering the characteristics of point clouds collected by Vehicle-Borne Laser Scanning System, a new method is put forward on how to automatically identify and locate poles from laser scanning data. The Density of Projected Points and the geometrical quality of poles are taken into account to automatically identify and locate poles.And experimental results prove that the method is feasible and available.6. An automatic road marking detection algorithm from LiDAR point clouds is proposed. Combining LiDAR features of retro, angle and distance with the properties of traffic marking, road marking is extracted, fitted and then used to generate CAD map.