Accuracy Improvement For Mobile Mapping In Complex Ground Environments

The multi-sensor integration based mobile mapping system (MMS) is an e-merging, efficient and convenient means for spatial information acquisition. With the support of modern sensor technologies, compute science, pattern recognition technologies and error adjustment theory, MMS is able to acquire multiple kinds of Geo-spatial data, such as vehicle trajectories, panoramas, points cloud and so on, efficiently at the same time. After automatically and intelligently processing and deeply mining the massive spatial data, the precise three dimensional model of surveyed objects and the digital-internet map, which are the basis of customer-oriented, event-oriented and management-oriented location based services (LBS), can be constructed.For all kinds of MMS, how to eliminate errors to improve the data accuracy is one of the core issues for theoretical researches and engineering applications. Unlike traditional surveying methods, MMS integrates various sensors, employs the dynamic surveying mode, so its error source and error feature are more complicated. The error of MMS not only comes from the measurement error of sensors, but also comes from the spatial-temporal alignment error of multi-sources data fusion. To deal with MMS errors, the randomness, tendency and environmental relativity of various errors should be analyzed in detail first. Then, different errors are processed with different ways according to their features.In general, the MMS error has three sources:the error of the positioning and orientation system (POS), the error of load sensors (such as laser scanners, cam-eras and so on), and the error of alignment between load sensors and POS. After calibration, the error of load sensors can be treated as minor noises. In many litera-tures and researches, the error of alignment is treated as a kind of systematic error and calibrated through self-calibration or control points based calibration methods. The POS error is the most difficult one to deal with, for it is closely related to the environments.The POS equipped by MMS consists of Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS). It provides location, velocity, orien-tation for MMS and load sensors. The long-term accuracy of POS relies on GNSS, of which the accuracy mainly depends on the environments. The influence of the complex real environments on POS poses three challenges on the engineering im-plementations of MMS:first of all, in good GNSS environments, such as sky or highways with few buildings, how to eliminate the error of the ellipsoidal height presented by POS and the error of the local Geoid to improve the accuracy MMS dynamic leveling; secondly, in unstable GNSS environments, such as city canyons and big water areas, how to improve the absolute accuracy of POS to assume its accuracy meet the requirement by high-accuracy mobile mapping; last, in under-ground spaces, where GNSS denied, the POS error will increase rapidly if only INS used, so how to constrain the INS error to achieve high accuracy of underground mobile mapping?To deal with the above three challenges, the thesis focuses on the accuracy im-provement of MMS in complex environments. Based on the detailed and deep anal-ysis of the MMS error model and the POS error feature, we propose a sliding least squares collocation (SLSC) method for accuracy-improvement of GNSS/INS dynam-ic leveling, a least squares collocation based method for accuracy-improvement of the mobile Light Detection And Ranging (LiDAR) system (MLS), and a LiDAR/IN-S/Odometer tightly coupled positioning and orientation method for underground high-accuracy mobile mapping, respectively. The methods proposed by this thesis are validated with data collected by various kinds of MMS under various environ-ments. The work of this thesis will help to break implementation constraints of MMS caused by complex environments, and supply the theoretic and experimental support for accuracy improvement of MMS.