Real-time robust vehicle detection through the same algorithm both day and night

Abstract

We propose a vehicle detection method based on traffic flow images obtained from a video camera set up on a low place such as the roadside or a sidewalk. We use the shadows underneath vehicles as a means of detecting them. Our method specifies the size of each vehicle according to the distance between its front and rear tires and our method also specifies the lane on which it exists. Our method has the advantage of automatically creating and updating background images and of automatically estimating and updating a threshold value we need to binarize background subtraction images and to enhance the vehicle detection accuracy. As a result, we have confirmed that our vehicle detection can be achieved by the same algorithm both day and night. The proposed algorithm is intended for a high-speed processing without complicated calculations and a real-time vehicle detection by using a general-purpose personal computer. Experimental results of our method in the four conditions, namely, in fine, cloudy, and rainy days and at nighttime, show that our vehicle detection accuracy is 96.5%.