Traffic News by Dynamic Fuzzy Classification

A necessary prerequisite for numerous services in traffic telematics is a good knowledge of the current traffic situation by location and time. An interdisciplinary solution is presented, which provides a location-time-dynamical description of the traffic situation by revealing traffic domains with uniform traffic states. For this pattern recognition task in a dynamical process with even chaotic traffic state transitions, the underlying measurements are classified into different traffic states. Some specific problems arising in this context are solved: 1) Data from different sources and of different units of measurement, e g., velocities, flows and densities, which are only sparsely available with respect to location and time are combined (“data fusion”). 2) “Floating Car Data (FCD)” is integrated using morphological filters and recognizing its location-time relation compared to other data. 3) The stochastic data are filtered without suppressing significant state transitions. 4) The subjective feeling of roadusers, that it is difficult to distinguish between different traffic states, is taken into account by fuzzy-classification. 5) Using a region growing method the segmentation problem of traffic domains is solved without being restricted to a predefined coarse road segmentation. 6) Good stability is obtained despite contradictory demands for a high resolution, a short reaction time and the differentiation of more than two traffic states. The chosen approach is confirmed by results with actual traffic data. The author knows of no other procedure with comparable capabilities.