Font Size: 

The paper presents a method of vehicles shape description using fuzzy rules. This method was intended for automatic class recognition of vehicles recorded in image data. Premises of the rules take into account selected shape coefficients determined for reference vehicles images (templates); consequences are associated with classes of vehicles. The rules induction procedure requires segments determination of the reference image. Fuzzy rules describe shape and arrangement of these segments that correspond to visible parts of a vehicle. A reference vehicle image for the rules induction is generated on the basis of three-dimensional shape model. A set of rules defined for different shape models enables vehicles classification. Input data of such fuzzy classifier consists of shape coefficients computed for segments recognised in a registered digital image of vehicle. Segments in the registered images are extracted using background updating, edge detection and area filling algorithms. The devised method is suitable for application in video sensors for road traffic control systems.