Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Survey on 3D Skeleton Based Person Re-Identification: Taxonomy, Advances, Challenges, and Interdisciplinary Prospects

Haocong Rao, Chunyan Miao

TL;DR

This survey addresses SRID by formalizing the task of identifying individuals from 3D skeleton sequences and tracing its evolution from hand-crafted descriptors to modern deep learning approaches. It proposes a taxonomy based on skeleton modeling (hand-crafted, sequence-based, graph-based) and learning paradigms (supervised, self-supervised, unsupervised), and surveys benchmarks, evaluation protocols, and model efficiencies. Empirical insights across public datasets and a case study on interdisciplinary applications illustrate SRID's current capabilities and limitations. The authors highlight data scarcity, noise, robustness, and generalization as key challenges and propose open directions, including multi-modal learning, cross-modality evaluation, skeleton foundation models, and privacy-conscious applications in healthcare, embodied AI, and security.

Abstract

Person re-identification via 3D skeletons is an important emerging research area that attracts increasing attention within the pattern recognition community. With distinctive advantages across various application scenarios, numerous 3D skeleton based person re-identification (SRID) methods with diverse skeleton modeling and learning paradigms have been proposed in recent years. In this paper, we provide a comprehensive review and analysis of recent SRID advances. First of all, we define the SRID task and provide an overview of its origin and major advancements. Secondly, we formulate a systematic taxonomy that organizes existing methods into three categories centered on hand-crafted, sequence-based, and graph-based modeling. Then, we elaborate on the representative models along these three types with an illustration of foundational mechanisms. Meanwhile, we provide an overview of mainstream supervised, self-supervised, and unsupervised SRID learning paradigms and corresponding common methods. A thorough evaluation of state-of-the-art SRID methods is further conducted over various types of benchmarks and protocols to compare their effectiveness, efficiency, and key properties. Finally, we present the key challenges and prospects to advance future research, and highlight interdisciplinary applications of SRID with a case study.

A Survey on 3D Skeleton Based Person Re-Identification: Taxonomy, Advances, Challenges, and Interdisciplinary Prospects

TL;DR

This survey addresses SRID by formalizing the task of identifying individuals from 3D skeleton sequences and tracing its evolution from hand-crafted descriptors to modern deep learning approaches. It proposes a taxonomy based on skeleton modeling (hand-crafted, sequence-based, graph-based) and learning paradigms (supervised, self-supervised, unsupervised), and surveys benchmarks, evaluation protocols, and model efficiencies. Empirical insights across public datasets and a case study on interdisciplinary applications illustrate SRID's current capabilities and limitations. The authors highlight data scarcity, noise, robustness, and generalization as key challenges and propose open directions, including multi-modal learning, cross-modality evaluation, skeleton foundation models, and privacy-conscious applications in healthcare, embodied AI, and security.

Abstract

Person re-identification via 3D skeletons is an important emerging research area that attracts increasing attention within the pattern recognition community. With distinctive advantages across various application scenarios, numerous 3D skeleton based person re-identification (SRID) methods with diverse skeleton modeling and learning paradigms have been proposed in recent years. In this paper, we provide a comprehensive review and analysis of recent SRID advances. First of all, we define the SRID task and provide an overview of its origin and major advancements. Secondly, we formulate a systematic taxonomy that organizes existing methods into three categories centered on hand-crafted, sequence-based, and graph-based modeling. Then, we elaborate on the representative models along these three types with an illustration of foundational mechanisms. Meanwhile, we provide an overview of mainstream supervised, self-supervised, and unsupervised SRID learning paradigms and corresponding common methods. A thorough evaluation of state-of-the-art SRID methods is further conducted over various types of benchmarks and protocols to compare their effectiveness, efficiency, and key properties. Finally, we present the key challenges and prospects to advance future research, and highlight interdisciplinary applications of SRID with a case study.
Paper Structure (15 sections, 12 equations, 4 figures, 4 tables)

This paper contains 15 sections, 12 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Overview
  3. Task Description
  4. Origin and Advancements
  5. Taxonomy of SRID Methods
  6. Skeleton Modeling
  7. Hand-Crafted Modeling
  8. Sequence-Based Modeling
  9. Graph-Based Modeling
  10. Learning Paradigms
  11. Benchmarks and Evaluation
  12. Challenges and Prospects
  13. Emerging and Open Directions
  14. Interdisciplinary Prospects
  15. Conclusion

Figures (4)

  • Figure 1: Overview of 3D skeleton based person re-ID (SRID) task with hand-crafted, sequence-based or graph-based modeling to learn effective body and motion features for identity recognition.
  • Figure 2: (a) Overview of research origin and technical advancements of SRID within the person re-ID community (Zoom in and follow the timeline for the best view). (b) Parameter sizes (Millions (M)), computational complexity (Giga Floating Point Operations (GFLOPs)), and KS20 Rank-1 accuracy of state-of-the-art deep learning methods for SRID (Red: Sequence-based models; Green: Graph-based models).
  • Figure 3: Structure of this survey with the taxonomy of SRID research. Representative branches and SRID methods are listed.
  • Figure 4: Interdisciplinary application landscape of SRID across three primary domains: healthcare (green box), embodied AI (yellow box), and security (purple box). Please zoom in for better view.