Recognizing Identities From Human Skeletons: A Survey on 3D Skeleton Based Person Re-Identification

2024-01-27

AI Summary

This paper systematically investigates skeleton-based 3D person re-identification (SRID), establishing for the first time a unified taxonomy comprising handcrafted features, sequential modeling, and graph-structured modeling approaches. It comprehensively surveys supervised, self-supervised, and unsupervised learning paradigms, along with their underlying semantic modeling mechanisms. We propose a unified evaluation framework integrating skeleton dynamics modeling, temporal networks (LSTM/Transformer), graph convolutional networks (GCN), and self-supervised strategies—including contrastive learning and masked skeleton reconstruction—and conduct extensive benchmarking across NTU, PKU, and BIWI datasets under multiple protocols. Experiments demonstrate that synergistic graph modeling and self-supervision significantly enhance cross-view robustness. We identify current bottlenecks—particularly in cross-domain generalization and weakly supervised learning—as critical future directions. The work provides reproducible benchmarks and practical guidelines for method selection.

Technology Category

• Computer Vision: 3D Computer Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

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 survey, 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 based on different skeleton modeling ($i.e.,$ hand-crafted, sequence-based, graph-based). Then, we elaborate on the representative models along these three categories with an analysis of their merits and limitations. Meanwhile, we provide an in-depth review of mainstream supervised, self-supervised, and unsupervised SRID learning paradigms and corresponding skeleton semantics learning tasks. A thorough evaluation of state-of-the-art SRID methods is further conducted over various types of benchmarks and protocols to compare their effectiveness and efficiency. Finally, we discuss the challenges of existing studies along with promising directions for future research, highlighting research impacts and potential applications of SRID.

Problem

• 3D skeleton-based person re-identification
• Skeleton modeling and learning paradigms
• Evaluation of SRID methods

Innovation

• 3D skeleton modeling
• diverse learning paradigms
• systematic taxonomy organization