In this thesis a two-step strategy for monitoring the urban traffic by analyzing the single-pass ALS data is presented and investigated. Two methods are proposed for vehicle extraction, namely a morphological segmentation of gridded data and a 3D segmentation of point clouds by an adaptive “mean shift” approach. Depending on the studied scene either one of the methods is applied alone or the combination of two methods by means of grouping large objects via “normalized cuts” is used. Afterwards, the motion state is derived by using a classifier based on Lie group metrics and the velocity of moving vehicles is determined by inverting a model for motion artifacts. The validation of the strategy is achieved with real laser data sets of Toronto, Munich and Enschede, which have an average point density of five points per square meter.
«
In this thesis a two-step strategy for monitoring the urban traffic by analyzing the single-pass ALS data is presented and investigated. Two methods are proposed for vehicle extraction, namely a morphological segmentation of gridded data and a 3D segmentation of point clouds by an adaptive “mean shift” approach. Depending on the studied scene either one of the methods is applied alone or the combination of two methods by means of grouping large objects via “normalized cuts” is used. Afterwards,...
»