Abstract -This paper investigates vehicle detection, tracking and classification. The detection stage is based on the background subtraction technique, where the background image is subtracted from the current frame. Pixel values greater than the set threshold are considered as part of the foreground. The tracking stage is accomplished using the Kalman ﬁlter, which predicts the current position of the vehicle using information of its prior positions. The tracked vehicle is then segmented at point where its color is most visible (point closest to the camera). A small patch is then extracted from the segmented vehicle and use as a color sample. At the final stage, the extracted patch is compared against the gallery to determine the color class of the vehicle. The experimental results show a promising trend in classifying vehicles based on their color with 100% accuracy in terms of classification.
No. of Downloads :