Fast algorithm for pedestrian and group of pedestrians detection using a laser scanner

Because they have neither well defined shapes nor well defined behaviors, detecting, tracking and classifying pedestrians in a dense urban environment from a moving vehicle remains a difficult task. This is especially true when people are standing or walking very close from one another. Indeed, because of occlusions, pedestrians are then usually very difficult to discriminate and several pedestrians can be wrongly detected as one unique obstacle leading ultimately to misclassifications. As a result, a great number of vulnerable people are likely to be missed. We present in this paper an algorithm that not only detect and track regular pedestrians but also cope smoothly and efficiently with groups of people. An original features based classification approach is also introduced. This algorithm, designed to be a part of an onboard collision avoidance system, meets two important requirements: it is fast and robust as proved by the experimental results presented in this paper.