ResearchMobility
Verkehrsregulator Erkennung und Identifizierung aus Crowdsourced Daten

Traffic Regulator Detection and Identification from Crowdsourced Data

Team:  Zourlidou, Sester
Year:  2020

Mapping with surveying equipment is a time-consuming and cost-intensive procedure that makes the frequent map updating unaffordable. In the last few years, much research has focused on eliminating such problems by counting on crowdsourced data, such as GPS traces. An important source of information in maps, especially under the consideration of forthcoming self-driving vehicles, is the traffic regulators. This information is largely lacking in maps like OpenstreetMap (OSM) and this research is motivated by this fact. 

The vehicular traffic at an intersection is regulated through a set of traffic rules or controls: either in form of traffic signals or traffic signs/rules (stop, yield/priority signs) that influence driving decision processes according to the current traffic. At such regulated areas, the observable movement of vehicles is affected by these rules, with the most commonly observed movement pattern being the moderation of the driving speed. Under this observation, the objective of identifying the regulation type of a junction is explored as a learning task. In particular, a learning task for classifying junctions by applying machine learning techniques on opportunistically collected vehicle trajectories. Various movement-based features, as well as statistical and map-based features are computed and then being used as input to train classifiers for identifying the regulation control system of juctions.   

Main reaseach aspects of this topic are (1) to identify the range of detected and recognised regulatory types by crowdsensing means, (2) to indicate the different classification techniques that can be used for these two tasks, (3) to assess the performance of different methods, as well as (4) to identify important aspects of the applicability of these methods.

As a proposed future development, GPS-based traffic regulation inference could be opportunistically “assisted” from imagery data when, for example, the classification accuracy in certain locations is either low or when junctions are sparsely sampled by GPS tracks. The latter is a common problem observed in most datasets as some junctions are more “popular” than others and are crossed by more vehicles and others are visited rarer, resulting in imbalances of the regulator types of the dataset. In such cases, a vision-based approach such as traffic-sign recognition of street-level images could be used to clarify the junction regulator context.