Overhead LIDAR-based motorcycle counting

In this paper, the authors investigate the problem of detecting and counting motorcycles in real traffic flow with a single layer laser scanner. This work focuses on the use of simple signal processing techniques to achieve a good performance in real time. The reported results outperform state-of-the art results. First, the authors report a 99.2% correct classification rate on 16 min sequence for a three-vehicle class classification problem. Second, the authors achieve a 0.8% counting error rate on 2 h of real traffic.