Hypergraphs for Joint Multi-view Reconstruction and Multi-object Tracking

Abstract

We generalize the network flow formulation for multiobject tracking to multi-camera setups. In the past, reconstruction of multi-camera data was done as a separate extension. In this work, we present a combined maximum a posteriori (MAP) formulation, which jointly models multicamera reconstruction as well as global temporal data association. A flow graph is constructed, which tracks objects in 3D world space. The multi-camera reconstruction can be efficiently incorporated as additional constraints on the flow graph without making the graph unnecessarily large. The final graph is efficiently solved using binary linear programming. On the PETS 2009 dataset we achieve results that significantly exceed the current state of the art.

Tracking Multiple-person using Sparse Stereo Information

“…It has been proved that HJRMT achieved results which significantly exceed the current state of the art by testing on the PETS 2009 dataset, even if detections from only one view were used [7]. In this paper, comparisons between our method and HJRMT of Scene 1 are given (object detector in Scene 2 performs poorly due to the large oblique angle).…”

“…For comparison, we implemented HJRMT (Hypergraphs for Joint Multi-View Reconstruction and Multi-Object Tracking) [7]. It has been proved that HJRMT achieved results which significantly exceed the current state of the art by testing on the PETS 2009 dataset, even if detections from only one view were used [7].…”

“…If the angle ( Figure 1) changes a lot, the detector has to be retrained before being put into use. For object tracking, a class of successful approach defines multiple people tracking to be a global optimization problem over the complete sequence [5][6][7] and shows good results. However, it is highly depends on the performance of the pedestrian detector, and it is definitely very time-consuming because all the detection results of the pedestrian detector have to be input through the whole sequence.…”

Inter-Person Occlusion Handling with Social Interaction for Online Multi-Pedestrian Tracking

“…Grouping, people intend to walk together, is another kind of social interaction that occurs frequently in the tracking scenario [16]. Because this work performs tracking online, we cannot access the complete pedestrians' trajectories to derive global groups, as in offline tracking methode [7], [17]. In this work, we instead infer a simple framewise pairing grouping relation.…”

Monocular Multiview Object Tracking with 3D Aspect Parts

“…An alternative way to achieve multiview object tracking is to utilize multi-camera settings, where the target is observed from multiple cameras simultaneously [21,23,17]. Tasks such as occlusion reasoning [21] and 3D reconstruction [17] which are challenging in monocular settings can be solved efficiently in multi-camera environments. Since multiple cameras are only available in specific scenarios, we focus on monocular multiview tracking in this work.…”