As smart city construction enters a deep-water phase, a core proposition has grown increasingly clear: who will guard the identity boundary of the digital city?
Over the past decade, facial recognition has handled most identity verification tasks in public security scenarios. However, as attack methods escalate — with deepfake technology proliferating and frequent cases of high-definition photos bypassing liveness detection — the single facial recognition system is exposing its structural vulnerabilities.
Against this backdrop, iris recognition is undergoing a strategic shift from "a tool for high-security special scenarios" to "a core component of city-level security infrastructure".
In May 2026, the latest IREX 10 ranking released by NIST shows that the False Match Rate (FMR) of top-tier iris recognition algorithms can be controlled below one in a million (10⁻⁶), far outperforming fingerprint recognition (typically at the one-in-ten-thousand level) and reaching the lowest level among current biometric technologies. The global Biometrics as a Service (BaaS) market reached $4.2 billion in 2025 and is expected to exceed $5 billion in 2026.
I. Why Iris? Technical Barriers Define Application Boundaries
Among all mainstream biometric technologies, iris recognition boasts three irreplaceable core advantages:
Lifelong Stability. Iris texture forms 6–18 months after birth and remains unchanged for life, unaffected by disease, aging, or labor wear. By contrast, fingerprints degrade with age and occupational environments, while facial features shift significantly with weight fluctuations, surgeries, and aging.
Exceptional Randomness and Uniqueness. The human iris contains approximately 266 independent quantifiable feature points, with far higher randomness than fingerprints (around 35). Even identical twins have completely distinct iris textures, and the left and right eyes of the same person are also independent and uncorrelated. This property gives iris recognition a theoretical false match rate on the order of 10⁻¹² — the lowest among all biometric technologies today.
Contactless Operation with Robust Liveness Detection. Iris capture requires no physical contact, making it naturally suited to scenarios with strict public health requirements. Combined with near-infrared illumination and multi-dimensional liveness detection, it can effectively defend against attacks using photos, videos, and 3D-printed prosthetic eyes.
These three characteristics are the fundamental reason why iris recognition is irreplaceable in border port control, judicial security, medical identity verification, and high-net-worth financial scenarios.
II. Phaselirs™: Core Competitiveness at the Algorithm Level
The commercialization threshold for iris recognition lies not in hardware, but in algorithms — more precisely, in the ability to extract reliable iris features at high speed under real-world conditions such as complex lighting, motion blur, and partial occlusion.
The self-developed Phaselirs™ phase-based iris recognition algorithm from Homsh is a technical barrier built to address this core pain point. Its core innovation lies in the introduction of a phase-domain feature encoding framework:
● Multi-scale Gabor phase analysis: Extracts iris texture features in the frequency domain rather than the spatial domain, delivering inherent robustness to image blur and noise;
● Adaptive boundary segmentation engine: Achieves stable iris localization even with strong pupillary reflection or up to 40% eyelid occlusion;
● Dynamic quality assessment module: Scores captured frames in real time and automatically selects the optimal frame for matching, reducing on-site capture failure rates to below 1%.
In real-world deployment verification, Phaselirs™ delivers millisecond-level single matching speeds for 1:N retrieval against a million-level database, meeting the throughput demands of city-level high-concurrency scenarios.
III. Qianxin + OVAI: An End-to-End System Closed Loop
Breakthroughs at the algorithm level can only unlock full commercial value when paired with deep hardware integration.
Homsh's upcoming Qianxin series of dedicated AI chips are deeply customized and optimized for the Phaselirs™ algorithm:
● Built-in dedicated Neural Processing Unit (NPU) for iris feature extraction, cutting per-frame processing energy consumption by more than 60%;
● Supports iris + face multimodal fusion inference, completing dual identity verification on a single chip without additional coprocessors;
● Industrial-grade wide-temperature design (-40°C ~ +85°C), suitable for outdoor smart city deployment scenarios.
Built on the Qianxin chip, the OVAI all-in-one platform integrates a near-infrared lighting module, binocular camera array, and edge inference engine, forming a full closed loop of "capture – processing – matching – decision-making": binocular synchronous capture (<0.3s), local Phaselirs™ matching (no cloud dependency), and encrypted result reporting (compliant with Classified Protection 2.0 requirements).
The core value of this end-to-end architecture is that it completely eliminates network latency risks and data transmission security hazards of cloud-based matching. For highly sensitive scenarios such as public security, military, and finance, local processing is not an option — it is a necessity.
IV. Market Landscape: The Policy Window Has Opened
From 2025 to 2026, multiple policy signals have emerged simultaneously, opening a critical window for the large-scale rollout of iris recognition.
Regulatory Level: The detailed implementation rules of the Personal Information Protection Law emphasize the principle of "data minimization" for biometric information, pushing the industry to shift toward technical solutions with high accuracy and low false capture rates. The Data Security Law requires localized storage of identity verification data for critical infrastructure, imposing tangible constraints on cloud-based solutions and benefiting localized iris solutions.
Standard Level: The national standard GB/T 41988-2022 Image Technical Requirements for Public Security Iris Recognition Applications has officially taken effect, removing compliance barriers for iris recognition to enter government procurement catalogs. The ongoing NIST IREX 10 evaluations and continuous algorithm improvements by global top vendors are also driving domestic manufacturers to accelerate iteration.
Market Level: Cumulative investment in China's smart city construction during the 14th Five-Year Plan period has exceeded 4 trillion yuan, with the share of security infrastructure continuing to rise. Vendors that take the lead in completing vertical integration of algorithm, chip, and complete device will hold an irreplicable first-mover advantage.
When an AI-generated face can fool a camera, urban security systems need a more trustworthy anchor. That anchor is provided by the lifelong, unchanging iris texture — not because it is more fashionable, but because it is more fundamental.
From the Phaselirs™ algorithm to Qianxin chips, from OVAI terminals to system integration capabilities, what Homsh is building is a time-tested identity trust system — making iris recognition the most reliable security infrastructure for smart cities.
