Study On Automatic Vehicle Classification Based On Their Outline Geometrical Characters

Automatic vehicle classification (AVC) is the important prerequisite for automatic charging and administrating system. In this dissertation, many kinds of AVC system and the concrete vehicle type in our country are analysed. On the basis of that, the method of acquiring characteristics by infrared plane detection is given, according to the characteristics gained and the pattern recognition, a fuzzy-neural vehicle classifier based on the vehicle's length height and axles is carried out.In this dissertation, the newly enacted standard of vehicle classification for highway toll charging is adopted. According to the statistics of main vehicles running on the highway, a general and practical vehicletype database is created. On this basis, correlation analysis, singal-parameter analysis and double-parameter analysis are fulfilled. And through this, the characteristic parameters are decided. At last ,we use a fuzzy-neural arithmetic model (NEFCLASS) to create a vehicle classifier from learning the vehicle datas, obtain the fuzzy rules and fuzzy sets.The experiment shows, through the AVC system we designed, the outline geometrical characters of vehicles can be rapidly obtained, and the classifier can generally classify the passing vehicles. The result is mainly up to the mustard.