arXiv subjects

Stephen Simpson

Publications and source records attributed to Stephen Simpson.

2 recordsLinked to original sources

arXiv2026

KnifeHunter: Structured Local Representation Learning for Fine-Grained Knife Image Retrieval in Law Enforcement

Knife-enabled violence presents a major public safety challenge, and law enforcement agencies require scalable tools for catalogue-level knife identification, intelligence analysis, and source attribution. Manual visual comparison is specialist, time-consuming, and difficult to scale under operational imaging conditions. We introduce KnifeHunter, an end-to-end forensic knife image retrieval system developed with UK law enforcement. The work contributes the KnifeHunter dataset, comprising 25,843 images across 543 knife classes from police evidence, retail catalogues, and border-force seizures, with structured metadata, Medium/Hard evaluation protocols, and large-scale distractor evaluation. We further propose CoRe-Net, a compact single-descriptor retrieval architecture that combines global context with spatially localised discriminative evidence. CoRe-Net introduces Structured Complementary Representation Learning (SCRL) to organise local evidence into complementary prototype-based representations, and Bi-Directional Reciprocal Fusion (BDRF) to integrate global and local evidence through residual projection and gated local-to-global injection. Using an EVA02-Base backbone and cosine-similarity retrieval, CoRe-Net achieves 88.0% mAP and 86.7% mP@10 on the Medium protocol, and 85.1% mAP and 83.8% mP@10 under distractor conditions. KnifeHunter was deployed by UK police forces during Operation Sceptre deployments from 2023 to 2025, achieving 99.2% mP@1 on field queries. These results demonstrate a practical and effective multimedia retrieval framework for fine-grained forensic knife matching in operational law-enforcement settings.

cs.CV↗

arXiv2025

Extraction of Bend-Resolved Modal Basis in Deformed Multimode Fiber

Mode mixing in optical fibers caused by mechanical bending induces perturbations that distort the spatial field profile of coherent beams as they propagate through few-mode or multimode fibers. The observed output from a bent fiber commonly appears as complex speckle, which is challenging to relate directly to the underlying deformation, particularly in continuously varying systems such as aerially deployed fibers or fiber-integrated sensors in mechanical structures. We introduce a novel method for constructing a complete deformation-resolved orthonormal modal basis that captures the optical response of a multimode fiber across a range of controlled mechanical deformations. The basis is derived via a two-stage singular value decomposition framework that initially constructs deformation-specific orthonormal mode sets from speckle pattern correlation matrices and subsequently decomposes the aggregated sets to produce a unified functional basis that comprehensively spans the deformation-induced modal subspace supported by the fiber. This hierarchical framework yields an energy-balanced representation that isolates statistically dominant field components across all deformation states, approximates superpositions of the fiber's propagation-invariant modes, systematically encodes deformation-induced perturbations, and supports robust decomposition of output fields across varying mechanical conditions. Such a basis enables tracking of mechanically induced modal evolution in deployed fibers, supporting distributed sensing, network resilience, and predictive fault diagnostics, with potential for integration into mode-division multiplexing systems.

physics.optics↗