The Design and Implementation of Real-Time Automatic Vehicle Detection and Counting System

This paper designed and implemented with Visual C++ a software video stream processing system to realize the real-time automatic vehicle speed detection and vehicle number counting. Based on geometric optics, we first present a simplified method to accurately map the coordinates in image domain into real-world domain. The second part is focused on the vehicle detection in digital image frames in video stream. We tested this system on a laptop powered by an Intel Centrino-2 (1.2 GHZ) CPU and 1GB RAM. The processing speed is 18.0 frames/second. Experiment also shows the system is able to simultaneously perform vehicle speed detection and vehicle number counting over multiple lanes. The detected vehicle speed's average error is 2.3 km/hour and counting accuracy achieves 97.8% (2298 among 2340 vehicles) for a video clip of 42 minutes.