Multi-target tracklet stitching through network flows

Castañón, Gregory; Finn, Lucas

doi:10.1109/aero.2011.5747436

Cited by 27 publications

(7 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is worth mentioning that we adopt the network-flow formulation to model the data association subproblem in this paper. While there are many other variants of this type of modeling in literature [25,33,34,35,36], we are only interested in its original form here [2] for its low computational costs. We extend the model to account for the multi-category case where multiple layers of the network need to be constructed with a set of new linear constraints.…”

Section: Relation To the Proposed Methodsmentioning

confidence: 99%

“…Despite efforts to detect partially occluded objects, missed detection/false positives are still inevitable, and any ambiguity could be resolved in the data association stage. Research efforts for multiple object tracking typically treat occluded objects as missed detection events or track occluded objects all together with a single tracker and iteratively grow or stitch tracklets (track fragments) before and after occlusions [19,20,21,22,23,24,25,26,27,28]. The main effort in this research direction is to design a discriminant similarity measure for the tracklet matching problem so that occlusion is implicitly resolved by filling gaps after stitching tracklets.…”

Section: Occlusion Reasoning In Trackingmentioning

confidence: 99%

“…Other variants of the network flow model exist in the tracking literature[25,33,34], all of which were designed for single-category object tracking. Here we extend the classic network flow model to a multi-layer network model in order to handle the multi-object multi-category tracking problem.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Global optimization for coupled detection and data association in multiple object tracking

Betke

2016

Computer Vision and Image Understanding

View full text Add to dashboard Cite

We present a novel framework for tracking multiple objects imaged from one or more static cameras, where the problems of object detection and data association are expressed by a single objective function. Particularly, we combine a sparsity-driven detector with the network-flow data association technique. The framework follows the Lagrange dual decomposition strategy, taking advantage of the often complementary nature of the two subproblems. Our coupling formulation avoids the problem of error propagation from which traditional "detection-tracking approaches" to multiple object tracking suffer. We also eschew common heuristics such as "non-maximum suppression" of hypotheses by modeling the joint image likelihood as opposed to applying independent likelihood assumptions. Our coupling algorithm is guaranteed to converge and can resolve the ambiguities in track maintenance due to frequent occlusion and indistinguishable appearance between objects. Furthermore, our method does not have severe scalability issues but can process hundreds of frames at the same time. Our experiments involve challenging, notably distinct datasets and demonstrate that our method can achieve results comparable to or better than those of state-of-art approaches.

show abstract

Section: Relation To the Proposed Methodsmentioning

confidence: 99%

Section: Occlusion Reasoning In Trackingmentioning

confidence: 99%

See 1 more Smart Citation

Global optimization for coupled detection and data association in multiple object tracking

Betke

2016

Computer Vision and Image Understanding

View full text Add to dashboard Cite

show abstract

“…Recently, there has been a growing interest in reversible data association methods, in which association decisions can be revised as more information becomes available. However, because of the computational complexity typically only a single estimate over the association decisions -typically the maximum a posteriori one -is maintained [4].…”

Section: Introductionmentioning

confidence: 99%

Distributed Fusion of PHD Filters Via Exponential Mixture Densities

Üney

Clark

Julier

2013

IEEE J. Sel. Top. Signal Process.

187

130

View full text Add to dashboard Cite

In this paper, we consider the problem of Distributed Multi-sensor Multi-target Tracking (DMMT) for networked fusion systems. Many existing approaches for DMMT use multiple hypothesis tracking and track-to-track fusion. However, there are two difficulties with these approaches. First, the computational costs of these algorithms can scale factorially with the number of hypotheses. Second, consistent optimal fusion, which does not double count information, can only be guaranteed for highly constrained network architectures which largely undermine the benefits of distributed fusion. In this paper, we develop a consistent approach for DMMT by combining a generalised version of Covariance Intersection, based on Exponential Mixture Densities (EMDs), with Random Finite Sets (RFS). We first derive explicit formulae for the use of EMDs with RFSs. From this, we develop expressions for the probability hypothesis density filters. This approach supports DMMT in arbitrary network topologies through local communications and computations. We implement this approach using Sequential Monte Carlo techniques and demonstrate its performance in simulations.

show abstract

“…The association matrices can be solved by various algorithms including the auction algorithm [4], the integer programming algorithm [5], or by using an N -scan algorithm along with belief propagation [3]. Track fragments can also be modelled as nodes in a flow networks such as in [6] and [7]. The flow networks are then solved by associating the nodes with neighbouring nodes that are highly correlated.…”

Section: Introductionmentioning

confidence: 99%

Comparative investigation into Viterbi based and multiple hypothesis based track stitching

Merwe

Villiers²

2016

IET Radar, Sonar & Navigation

View full text Add to dashboard Cite

Abstract-In this study a comparative analysis is performed between a novel Viterbi based and multiple hypothesis based track stitching algorithms. The track fragments in the Viterbi based track stitching algorithm are modelled as nodes in a trellis structure. A sequential Viterbi data association algorithm is then used to solve the trellis and associate track fragments with each other. A Kalman filter is used to determine the possible associations as well as the probabilities of the associations between the track fragments. In the multiple hypothesis track stitching algorithm, the hypothesis based multiple hypothesis tracking (MHT) algorithm is extended to perform track fragment to track fragment associations, rather than associating observations to tracks. Aspects of the developed multiple hypothesis algorithm are compared with implementations of a similar nature. Novel aspects of this research include the modification of the sequential Viterbi algorithm, as well as the extension of the MHT algorithm to solve the track stitching problem. It was found that the sequential Viterbi track stitching algorithm performed somewhat better than the multiple hypothesis track stitching algorithm for similar execution times. The Viterbi based track stitching algorithm is also shown to produce more consistently acceptable results.

show abstract

Multi-target tracklet stitching through network flows

Cited by 27 publications

References 10 publications

Global optimization for coupled detection and data association in multiple object tracking

Global optimization for coupled detection and data association in multiple object tracking

Distributed Fusion of PHD Filters Via Exponential Mixture Densities

Comparative investigation into Viterbi based and multiple hypothesis based track stitching

Contact Info

Product

Resources

About