A Real Time Window Based Machine Vision System for Counting Number of Vehicles on a Street Lane

In this paper we propose a real time image processing technique for counting the number of vehicles passing along a street lane. In this method, real time speed is achieved by processing, at each frame, only the pixels inside a defined small window across a lane. For detecting vehicles, several processes are done including, enhancement, background differencing and mask based edge detection. After these processes a binary image showing the vehicle edges is created inside the small window. Extracting the number of vehicles is achieved by analyzing the number of edge pixels detected inside the small window for a specific time interval. In addition, we have used a real time image enhancement method to overcome the problem of variation in ambient light levels during different hours of day. Our method was tested in different traffic situations and ambient light levels. The results of vehicle counting were compared with that done by a human expert. Satisfactory results with accuracies over %94.8 were obtained.