Behaviour Reasoning and Tracking for Mobile Objects Based on Projection and Occlusion Ratios

Object states estimation and data association are main facets of multi-object tracking. Under complex situations, one object often grouped with others, or occluded by other objects or background, which can increase the difficulty in data association and behaviour reasoning, and even result in being hard to maintain track. To cope with these problems, a new behaviour reasoning and tracking method is proposed. A scene mode is built firstly simulating public transport environment. Then, Kalman filter is used to estimate the object position and bounding box. Then, the center-association based projection ratio and region-association based occlusion ratio are defined and combined to reason and classify object behaviours. Finally, an adaptive tracking scheme aiming at special behaviours (e.g. merging, and static occlusion) is proposed to validate our reasoning method. Experimental results show that the proposed method is efficient.