PKF: Probabilistic Data Association Kalman Filter for Multi-Object Tracking

2024-11-10

arXiv.org

AI Summary

To address accuracy degradation and computational overhead caused by measurement-to-state association uncertainty in multi-object tracking (MOT), this paper proposes the Probabilistic Data Association Kalman Filter (PKF). Methodologically, PKF treats data association as a latent variable and employs variational inference with an EM algorithm for efficient Bayesian state estimation. It introduces the first explicit use of the permanent of the likelihood matrix to compute association probabilities and incorporates an ambiguity-aware detection mechanism that applies probabilistic association only to ambiguous subsets—balancing accuracy and efficiency. Additionally, PKF adopts an extended measurement vector and a lightweight bounding-box association strategy. Experiments demonstrate that PKF achieves state-of-the-art HOTA scores among Kalman-based MOT methods on MOT17, MOT20, and DanceTrack, with CPU-only inference exceeding 250 fps. Remarkably, using detections alone—without appearance or velocity features—PKF ranks within the top-10 on both MOT17 and MOT20 leaderboards.

Technology Category

• Intelligent Robots: State Estimation
• Computer Vision: Motion & Tracking
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

In this paper, we derive a new Kalman filter with probabilistic data association between measurements and states. We formulate a variational inference problem to approximate the posterior density of the state conditioned on the measurement data. We view the unknown data association as a latent variable and apply Expectation Maximization (EM) to obtain a filter with update step in the same form as the Kalman filter but with expanded measurement vector of all potential associations. We show that the association probabilities can be computed as permanents of matrices with measurement likelihood entries. We also propose an ambiguity check that associates only a subset of ambiguous measurements and states probabilistically, thus reducing the association time and preventing low-probability measurements from harming the estimation accuracy. Experiments in simulation show that our filter achieves lower tracking errors than the well-established joint probabilistic data association filter (JPDAF), while running at comparable rate. We also demonstrate the effectiveness of our filter in multi-object tracking (MOT) on multiple real-world datasets, including MOT17, MOT20, and DanceTrack. We achieve better higher order tracking accuracy (HOTA) than previous Kalman-filter methods and remain real-time. Associating only bounding boxes without deep features or velocities, our method ranks top-10 on both MOT17 and MOT20 in terms of HOTA. Given offline detections, our algorithm tracks at 250+ fps on a single laptop CPU. Code is available at https://github.com/hwcao17/pkf.

Problem

• Develops probabilistic data association Kalman filter for multi-object tracking
• Approximates state posterior density using variational inference and EM
• Reduces association time while maintaining tracking accuracy

Innovation

• Uses probabilistic data association Kalman filter
• Applies Expectation Maximization for latent variables
• Reduces association time with ambiguity checks