2102 インフラレーザスキャナによる人物追跡に関する研究 : グループを構成する複数人物の追跡(OS10-1 ビークルオートメーション,OS10 アドバンティシンポジウム,オーガナイズド・セッション(OS))

Abstract

This paper presents a laser-based tracking of people in a group. Our method consists of people-detecting and people-tracking procedures. From the data taken by the two-layered laser scanner (LS), the position data of people is taken by background subtraction. By using the position data of people, heuristic-rule-based and global-nearest-neighbor (GNN)-based data association identifies multiple people in crowded environments. Their identified people are tracked via model-based tracker; the interacting-multi-model(IMM) estimator is applied to tracking people with sudden changes in motion; walking, running or stopping. People located in the vicinity with similar motions are grouped. When people are occluded by other people in the same group, the occluded people are tracked based on laser measurements related to the nearest-neighbor person in the group. Then the tracking lost can be reduced and the tracking performance can be improved. Experimental result validates the effectiveness of our method.