Towards reliable real-time multiview tracking

2002

In this paper we address the problem of reliable real-time 3D-tracking of multiple objects which are observed in multiple wide-baseline camera views. Establishing the spatio-temporal correspondence is a problem with combinatorial complexity in the number of objects and views. In addition vision-based tracking suffers from the ambiguities introduced by occlusion, clutter and irregular 3D motion.

Pattern Recognition Letters, 1998

A method of tracking multiple objects of known geometry using multiple cameras is proposed. Our approach differs from the previous approaches in that the object geometry is tightly integrated into the tracking process. The major contribution is threefold: Firstly, multiple cameras are used to improve the accuracy of the estimated posture parameters. Additional formalism required by considering multiple images is nicely integrated into the tracking model, and is handled effectively. Secondly, the feature tracking is facilitated by integrating the measurement and dynamic models into the matching process, thereby improving the accuracy and robustness of the feature correspondence. Thirdly, ambiguities that may arise in the course of the feature matching are resolved by the statistical analysis and the visibility test. The entire process from the image sequence to the posture parameters has been completely automated into a single, seamless process, and has been extensively tested on synthetic and real images.

Object Tracking, 2011

First ACM SIGMM international workshop on Video surveillance - IWVS '03, 2003

This paper presents novel approaches for continuous detection and tracking of moving objects observed by multiple, stationary or moving cameras. Stationary video streams are registered using a ground plane homography and the trajectories derived by Tensor Voting formalism are integrated across cameras by a spatio-temporal homography. Tensor Voting based tracking approach provides smooth and continuous trajectories and bounding boxes, ensuring minimum registration error. In the more general case of moving cameras, we present an approach for integrating objects trajectories across cameras by simultaneous processing of video streams. The detection of moving objects from moving camera is performed by defining an adaptive background model that uses an affine-based camera motion approximation. Relative motion between cameras is approximated by a combination of affine and perspective transform while objects' dynamics are modeled by a Kalman Filter. Shape and appearance of moving objects are also taken into account using a probabilistic framework. The maximization of the joint probability model allows tracking moving objects across the cameras. We demonstrate the performances of the proposed approaches on several video surveillance sequences.

ArXiv, 2020

This paper proposes an online multi-camera multi-object tracker that only requires monocular detector training, independent of the multi-camera configurations, allowing seamless extension/deletion of cameras without (retraining) effort. The proposed algorithm has a linear complexity in the total number of detections across the cameras, and hence scales gracefully with the number of cameras. It operates in 3D world frame, and provides 3D trajectory estimates of the objects. The key innovation is a high fidelity yet tractable 3D occlusion model, amenable to optimal Bayesian multi-view multi-object filtering, which seamlessly integrates, into a single Bayesian recursion, the sub-tasks of track management, state estimation, clutter rejection, and occlusion/misdetection handling. The proposed algorithm is evaluated on the latest WILDTRACKS dataset, and demonstrated to work in very crowded scenes on a new dataset.

2003

Conventional tracking approaches assume proximity in space, time and appearance of objects in successive observations. However, observations of objects are often widely separated in time and space when viewed from multiple non-overlapping cameras. To address this problem, we present a novel approach for establishing object correspondence across non-overlapping cameras. Our multi-camera tracking algorithm exploits the redundance in paths that people and cars tend to follow, e.g. roads, walk-ways or corridors, by using motion trends and appearance of objects, to establish correspondence. Our system does not require any inter-camera calibration, instead the system learns the camera topology and path probabilities of objects using Parzen windows, during a training phase. Once the training is complete, correspondences are assigned using the maximum a posteriori (MAP) estimation framework. The learned parameters are updated with changing trajectory patterns. Experiments with real world videos are reported, which validate the proposed approach.

The use of visual sensors may have high impact in applications where it ist required to measure the pose (position and orientation) and the visual features of object moving in unstructured environments. In robotics, the measurement provided by video cameras can be directly used to perform closed-loop control of the robot end-effector pose. In this chapter the problem of real-time estimation of the position and orientation of a moving object using a fixed stereo camera system is considered. An approach based on the use of the Extended Kalman Filter (EKF) combined with a 3D representation of the objects geometry based on Binary Space Partition (BSP) trees ist illustrated. The performance of the proposed visual tracking algorithm is experimentally tested in the case of an object moving in the visible space of a fixed stereo camera system.

IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020

This paper proposes an online multi-camera multi-object tracker that only requires monocular detector training, independent of the multi-camera configurations, allowing seamless extension/deletion of cameras without retraining effort. The proposed algorithm has a linear complexity in the total number of detections across the cameras, and hence scales gracefully with the number of cameras. It operates in the 3D world frame, and provides 3D trajectory estimates of the objects. The key innovation is a high fidelity yet tractable 3D occlusion model, amenable to optimal Bayesian multi-view multi-object filtering, which seamlessly integrates, into a single Bayesian recursion, the sub-tasks of track management, state estimation, clutter rejection, and occlusion/misdetection handling. The proposed algorithm is evaluated on the latest WILDTRACKS dataset, and demonstrated to work in very crowded scenes on a new dataset.

In this paper, we propose a new multiple-camera people tracking system that is equipped with the following functions: (1) can handle long-term occlusions, complete occlusions, and unpredictable motions; (2) can detect arbitrary sized foreground objects; (3) can detect objects with much faster speed. The main contribution of our method is twofold: 1) An Mto-one relationship with only point homography matching for occlusion detection can achieve efficiency; 2) A view-hopping technique based on object motion probability (OMP) is proposed to automatically select an appropriate observation view for tracking a human subject.

Motion and Video Computing, 2002. …, 2002

This paper presents a set of methods for multi view image tracking using a set of calibrated cameras. We demonstrate how effective the approach is for resolving occlusions and tracking objects between overlapping and non-overlapping camera views. Moving objects are ...