SVM-based detection of moving vehicles for automatic traffic monitoring

A traffic surveillant system must be capable of working in all kinds of weather and illumination conditions, such as shadows in a sunny day, vehicle reflections in a rainy day and vehicle headlights in the evening. In this paper we propose a robust algorithm to detect real moving vehicles and eliminate the influence of shadows and vehicle headlights by using a pattern classification method. On account of its simple but efficient representation, the histogram of a difference image is used as the features for classification. The classifier is designed based on support vector machine (SVM) due to its high generalization performance. The final experiment shows that the real traffic monitoring system based on our algorithm can detect moving vehicles and separate shadows and headlights robustly and effectively in different weather and illumination conditions.