Acquisition de l'alignement routier. Profil en long et trace en plan: A partir de traces GPS

A road alignment reconstitution tool has been developed; it analyses separately horizontal data and vertical data to define alignment (horizontal and vertical) characteristics. By using least squares regressions, distances between GPS points and recomposed curves of road alignment are minimized. In order to apply regressions, GPS databases must be treated. These treatments introduce control parameters which help to optimize the accuracy of reconstitution algorithm.; This algorithm based on least squares regressions fit the best curve minimizing distances. Other parameters intervene as constraints. A road alignment has to be continuous, and this implies that the different reconstituted curves and straight lines must be in tangency. This condition affects directly algorithm equations.; Efficiency of algorithm is evaluated with two approaches. First the algorithm accuracy is tested on fictive tracks with known geometric characteristics what is very useful for the algorithm calibration. It appears that the reconstitution tool gives very satisfactory results and that reconstituted alignments are very accurate. Second the algorithm is evaluated on databases provided by Quebec Ministry of Transportation. It is tested by measuring distances between GPS and recomposed curves. One more time, the tool proves alignments are very close to GPS tracks. Moreover the different developed control parameters help to improve or adjust the accuracy of the algorithm in function of examined purposes. (Abstract shortened by UMI.)...