Detection of Stops on Truck Trajectories Using Self-Organizing Maps

With the increase in the number of available Geo-location data, the number of possible application grows as well. There are thus multiple researches done to improve the detection of relevant places in trajectories. This is notably done through semantic enrichment of the data and through the use of new types of algorithm or improvements made on pre-existing methods. This thesis aims at detecting relevant logistic stops in truck trajectories. The present methodology uses the DBSCAN algorithm to split trajectories in multiple sub-trajectories which will is necessary to compute a new variable called move ability. The new variable added, I use the self-organizing map, an unsupervised form of neural network. Once the detection of the stops is done, I will compare the Belgian logistic network identified by the European commission with the network identified by the algorithm and thus assess the performance of the algorithm in identifying relevant logistic stops. In addition to this, I assessed the potential performance of land cover data as a potential semantic enrichment of trajectories to help in the detection of stops.