Robust vehicle detection even in poor visibility conditions using infrared thermal images and its application to road traffic flow monitoring

Abstract

We propose an algorithm for detecting vehicle positions and their movements by using thermal images obtained through an infrared thermography camera. The infrared thermography camera offers high contrast images even in poor visibility conditions like snow and thick fog. The proposed algorithm specifies the area of moving vehicles based on the standard deviations of pixel values along the time direction of spatio-temporal images. It also specifies vehicle positions by applying the pattern recognition algorithm which uses Haar-like features per frame of the images. Moreover, to increase the accuracy of vehicle detection, correction procedures for misrecognition of vehicles are employed. The results of our experiments at three different temperatures show that the information about both vehicle positions and their movements can be obtained by combining those two kinds of detection, and the vehicle detection accuracy is 96.2%. Moreover, the proposed algorithm detects the vehicles robustly in the 222 continuous frames taken in poor visibility conditions like snow and thick fog. As an application of the algorithm, we also propose a method for estimating traffic flow conditions based on the results obtained by the algorithm. By using the method for estimating traffic flow conditions, automatic traffic flow monitoring can be achieved.