Onboard real time system for 3D localization and modelling by multi-sensor data fusion
Iyad Abuhadrous · OpenGrey (Institut de l'Information Scientifique et Technique) · 2005
This Ph.D. thesis tackles the problem of localization of a moving vehicle using multisensor based approach. Therefore we will address the integration and the processing of data issued from heterogeneous navigation systems: GPS and inertial navigation system and odometry. The reason for this integration is to exploit the advantages of each system used. Integration uses a numerical filtering to proceed to a compensation of the errors in position, velocity, and attitude of each sensor. Emphasize was not address only on the integration, but also on the concepts of precision of the results. In the second time, we will be interested in the quantitative and qualitative validation of the results and the development of an application exploiting our work. The system errors are compared a reference trajectory. In a qualitative validation, to illustrate the efficiency of the system and as a direct application of it; we have developed a 3D environment digitizing system. A 3D modelling system of urban scenes was developed together a road modelling system. The system is based in the use of a single layer laser scanner for range data acquisition. A particular interest is related to the quality and the inaccuracies of the data returned by the scanner. Data scanning is performed with the flight and during the displacement of the vehicle. Finally, the last phase consists in exploiting the obtained model in two manners: the constitution of a precise geometrical and topological model then the extraction of semantic information of the model (extraction of the road or the fixed obstacles, the identifications of land-marks, the characterization of the road, etc.).