A vehicle classification system based on hierarchical multi-SVMs in crowded traffic scenes

Automatic vehicle classification is very important for video surveillance, especially for intelligent transportation system. Currently, some approaches have been proposed. However, almost all of these methods cannot play well in the practical crowded traffic scenes with heavy occlusions, shadows, and different views, etc. To solve this difficult problem, we propose a new vehicle classification method based on hierarchical multi-SVMs. First, we extract the foreground objects from the surveillance videos. Then, we use the proposed hierarchical multi-SVMs method for vehicle classification. Moreover, we present a voting based correction scheme by tracking the classified vehicles for the final precision. Based on the proposed approach, we have built a practical system for robust vehicle classification in complicated traffic scenes. Extensive experimental results show that our solution can achieve convincing results.