Model-based Multi-object Visual Tracking: Identification and Standard Model Limitations

2025-08-19

AI Summary

This paper addresses a systematic mismatch between the standard point-source object (SPO) model and pedestrian detection outputs (2D bounding boxes) in model-based multi-object visual tracking. To resolve this, we propose a bounding-box-observation-aware modeling refinement: the bounding box center is modeled as a continuous-time point target, integrated with a Poisson multi-Bernoulli mixture (PMBM) filter for posterior density estimation. The framework jointly infers target birth/survival probabilities, motion parameters, and observation uncertainty. Parameter learning and validation are conducted on the MOT-17 dataset, yielding competitive tracking performance. Crucially, this work quantitatively identifies— for the first time—the primary sources of bias induced by the SPO assumption in bounding-box tracking. By explicitly linking modeling assumptions to empirical detection characteristics, our analysis provides both theoretical grounding and empirical evidence for developing general-purpose tracking dynamic models better aligned with visual detection outputs.

Technology Category

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

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Tracking, profiling, and countermeasures against them

Abstract

This paper uses multi-object tracking methods known from the radar tracking community to address the problem of pedestrian tracking using 2D bounding box detections. The standard point-object (SPO) model is adopted, and the posterior density is computed using the Poisson multi-Bernoulli mixture (PMBM) filter. The selection of the model parameters rooted in continuous time is discussed, including the birth and survival probabilities. Some parameters are selected from the first principles, while others are identified from the data, which is, in this case, the publicly available MOT-17 dataset. Although the resulting PMBM algorithm yields promising results, a mismatch between the SPO model and the data is revealed. The model-based approach assumes that modifying the problematic components causing the SPO model-data mismatch will lead to better model-based algorithms in future developments.

Problem

• Addresses pedestrian tracking using 2D bounding box detections
• Reveals mismatch between standard point-object model and data
• Identifies model parameters from MOT-17 dataset and principles

Innovation

• PMBM filter for pedestrian tracking
• SPO model with continuous time parameters
• Model-data mismatch analysis for improvements