Abstarct: On road vehicle type recognition generally plays an important role in intelligent transportation applications, increasing security and ease of movement on roads. Already existing methods for vehicle type recognition are the ones that get the information required for recognition using different sensors, including laser, radar and vision sensors and study such information, but among all, vision baxsed methods have been more attractive due to reliability of results and low hardware and time costs. Vehicle type recognition can be done in two levels: Recognizing type baxsed on dimension and general appearance properties of vehicle (like classifying vehicles into classes of heavy, light and so forth) or more precisely recognizing vehicle make and model. In this thesis for vehicle type and color recognition sections videos of cars passing through a toll-gate are used. These kind of videos are less paid attention because of their limited field of view and difficulties of vehicle region of interest extraction. Type and model recognition of vehicles passing through toll-gates is necessary for security applications, statistical surveys, and automated toll collection and so on. A real-time algorithm is presented and the average recall and precision of the detection stage are 94.75% and 98.83% respectively and the accuracy of the vehicle general type and vehicle model recognition stages are 98% and 96.65% respectively.