An Approach to Vehicle Recognition Using Supervised Learning
Takeo Kato, Yoshiki Ninomiya · 1998
Introduction To enhancesafet y andtdk7 e#ciency, a driver assistssi systs and anaut7W(kQj vehicle syste are being developed. Intk7j syst7j( a vehicle should have a functkQ t recognizetc road environment such as a road lane, a preceding vehicle andanot6" obst6"(" Machine vision is one of tk road environment recognitec meto ds. Preceding vehiclerecognit)B is one of tk most import) t funct)7k t develop suchsystW7k However, it is a di#cult tcult machine vision because vehicles have various appearances, duet tek7 various shapes and colors. And appearance is changed when lightgh changes. Preceding vehiclerecognitj7 meto ds using verter - and horizonti edgedetk7"j" are reportj [1]--[3], t1] are based on tk fact tct tct are significant vertj)k and horizontr edges in tk di#erent#e images of backfaces of passenger vehicles such as a sedan, a coupe and ast6)(k wagon. In a real road environment however, various kinds of vehicles run as well as Manuscript received October 12, 1999