[論文レビュー] Extended Object Tracking: Introduction, Overview and Applications

Karl Granström, Marcus Baum+1 more|GoeScholar The Publication Server of the Georg-August-Universität Göttingen (Georg-August-Universität Göttingen)|Mar 14, 2016

Target Tracking and Data Fusion in Sensor Networks参考文献 199被引用数 224

ひとこと要約

拡張オブジェクト追跡の包括的な調査で、問題を定義し、形状と状態のモデリングを概説し、単一および多物体追跡手法とセンサー全体での応用をレビューする。

ABSTRACT

This article provides an elaborate overview of current research in extended object tracking. We provide a clear definition of the extended object tracking problem and discuss its delimitation to other types of object tracking. Next, different aspects of extended object modelling are extensively discussed. Subsequently, we give a tutorial introduction to two basic and well used extended object tracking approaches - the random matrix approach and the Kalman filter-based approach for star-convex shapes. The next part treats the tracking of multiple extended objects and elaborates how the large number of feasible association hypotheses can be tackled using both Random Finite Set (RFS) and Non-RFS multi-object trackers. The article concludes with a summary of current applications, where four example applications involving camera, X-band radar, light detection and ranging (lidar), red-green-blue-depth (RGB-D) sensors are highlighted.

研究の動機と目的

• Define extended object tracking and distinguish it from point and contour tracking.
• Survey object state, measurement, shape, and dynamics modelling for extended objects.
• Introduce and compare two foundational extended object tracking approaches (random matrix and Kalman filter-based star-convex shapes).
• Discuss multi-object extension and how to manage many association hypotheses via Random Finite Sets and non-RFS trackers.
• Highlight current applications across lidar, radar, camera, and RGB-D sensors and point to future trends.

提案手法

• Present formal definitions and taxonomy of tracking types (point, extended, group, multi-path).
• Detail object state composition including position, kinematics, and extent (shape/size/orientation).
• Explain measurement modelling options: set of reflection points (sprb), spatial Poisson point process (PPP), and binomial/other models, with their likelihood forms.
• Describe shape modelling levels (no shape, basic geometric shapes, arbitrary shapes) and corresponding measurement distributions.
• Outline dynamics modelling for object motion and how it integrates with extended object state estimation.
• Compare two core extended object tracking approaches: the random matrix model and the random hypersurface model (RHM) for star-convex shapes.
• Discuss approaches for multi-object tracking via Random Finite Sets (RFS) and non-RFS methods.

実験結果

リサーチクエスチョン

• RQ1 How should extended objects and their extents be modelled to accurately infer shape and motion from sparse, multi-sensor measurements?
• RQ2 What are the advantages and limitations of the random matrix model versus the random hypersurface model for extended object extent estimation?
• RQ3 How can multi-object tracking with extended objects be efficiently performed given combinatorial data association?
• RQ4 What are representative applications and sensor modalities that demonstrate extended object tracking in practice?

主な発見

• Extended object tracking handles objects occupying multiple sensor resolution cells and requires jointly estimating shape and motion.
• Two prominent modelling approaches are the random matrix model and the random hypersurface model for star-convex shapes.
• Measurement modelling can use point reflection models (sprb) with data association, a spatial Poisson point process (PPP) model, or binomial-based models, depending on the sensor and scenario.
• Shape modelling can vary from no shape to simple geometric shapes to arbitrary shapes, with trade-offs between accuracy and computational complexity.
• Tracking multiple extended objects necessitates managing a large space of data associations, for which Random Finite Sets and non-RFS trackers offer scalable solutions.
• The survey discusses applications across camera, X-band radar, lidar, and RGB-D sensors, illustrating practical deployment of extended object tracking.