SVM based people counting method in the corridor scene using a single-layer laser scanner

People counting plays an important role in public safety, building automation control and other data analysis like consumer behaviors, passenger management and so on. Among all the public places, the corridor is one of the most common scenes. Therefore, this paper proposes a people counting system which is applied in the corridor. The proposed system counts the amount of the passing pedestrians as well as detects the direction of each pedestrian. A mirror reflection device is designed to achieve a double-plane scanning by a single lidar, which doesn't only double the scanning data and make use of the wasted scanning angle, but also realize the direction detection by judging which scanning plane the pedestrian has passed first. A novel shape descriptor is used in the sliding window algorithm, and a support vector machine is trained as a classifier. In the proposed algorithm, head-shoulder feature in the pedestrian point cloud is detected to get the point cloud of every pedestrian. In the experiment, actual pedestrian data are collected to confirm the proposed system. The proposed method shows a better performance than the comparison methods.