A traffic flow measurement system at night by using image processing

In this paper, we propose a simple algorithm to calculate the number of passing cars at night by using an image processing sensor for digital black and white images with 256 tone levels. To recognize cars, we capture their head lamps. The reflection of the head lamps is one of the most troublesome factors in recognizing cars. The main problem in this paper is how to recognize cars under the influence of the reflection of the head lamps especially in rainy days. In general, the image of a head lamp is nearly circular and the reflection is long and narrow. On the difference of these forms, we can exclude the reflection in our proposed algorithms For real-time operation and simple calculation, we recognize the existence of cars using fifteen lines with 256 tone levels. In the experimental application on a road, the recognition rate of a real-time operation is more than 90%. Moreover, we will also explain briefly how to recognize passing cars for 24 hours.