Improving On-Road Vehicle Detection Performance by Combining Detection and Tracking Techniques

This paper presents the combination method of detection and tracking to improve the recall of on-road vehicle detection. The vehicle detection method is a kind of the object detection where the vehicle in front of a car is detected using a camera installed on the car. In the proposed method, the Viola Jones detection and the support vector machine technique are combined complementary. Further the Lucas Kanade tracking technique is introduced to gain the true positive detection when both detection techniques fail to detect the vehicle. The experimental results show that the proposed method achieves the recall of 0.97, the precision of 0.96 and the frame rate of 25.54 fps (frames per second).