Vision based classification and speed estimation of vehicles using forward camera

The number of vehicles today are increasing. This makes the roads congested and makes it difficult for monitoring systems to detect the vehicles. A good traffic monitoring system must be used to be able to count, detect, and classify moving vehicles. Vehicle classification and detection is an important task that can reduce the traffic congestion. In this paper, Support Vector Machine (SVM) classifier, one of the most popular techniques, will be used for vehicle classification and detection. HOG features from the collected images were used for training of the different SVM classifier models which are: Linear, Quadratic, Cubic and Gaussian. Speed estimation will also be applied in order to track the speed of the classified vehicles. Results through experimentation showed that the Cubic SVM classifier model with an accuracy of 94.29% produced the best output for detection and classification compared to the other SVM models. The speed of the vehicles classified by the SVM classifier were also successfully estimated and shown.