Enhancing automatic speed estimation systems performance using support vector machines

Speed cameras are systems which normally include a camera, a calibration system, speed estimation software, and a vehicle recognition system (which includes a vehicle type detector and a LPR system). These systems are normally used to detect and identify vehicles disobeying a speed limit in a road. While these systems are used widely in the world, their use in Iran is restricted due to various difficulties in using LPR systems for Farsi license plates. Also, to reduce the setup cost, in many traffic law enforcement systems low-cost low-resolution IR cameras are used, which increases the LPR complexity. In this paper, it has been showed that how using statistical learning methods (specifically SVMs) enhanced the automatic speed estimation systems performance by improving the vehicle identification via reading the license plate. The focus will be on Farsi license plates recognition. Obviously the results can be used in many potential applications such as automatic toll collection and traffic control in restricted areas.