Traffic censuses presently are the basis for a wide range of traffic planning, management and telematics applications. Current data collection is either permanently installed (on motorways) or done manually by students. Problems of the traditional methods are on the one hand high costs of installation and on the other hand inaccurate data. Therefore, a new traffic stream analysis system for an automatic, accurate, cheap, mobile and self-sufficient registration of vehicles was developed. The system records a video stream continuously in order to analyse number plates automatically. The output of the number plate recognition software is a list including number plates, date, time and pictures of all vehicles. This list is the basis for many different data interpretations (identification of vehicle classes, political districts and section control between two units...) which run almost automatically. The results show high data quality under several tested conditions (night, different traffic conditions, rain, direct solar radiation...). All projects which were carried out until now were satisfactorily completed and furthermore used to improve the already existing system. Further developments shall improve the quality of the registration quota (now 90-95 per cent, target is up to 99 per cent), enable the automatic classification of all vehicles and a higher automation of data analysis (A). For the covering abstract of the conference see ITRD E212343.
Abstract