https://doi.org/10.1093/nsr/nwag601
Nature’s game of hide and seek is entering the age of intelligent machines. A new article in
National Science Review traces how animal camouflage, advanced materials and
artificial intelligence are converging to reshape the contest between detecting an object and keeping it concealed.
Prepared by a team at the National University of Defense Technology in Changsha, China, the article brings together three domains:
biomimetics,
spectral physics and
AI. Its central message is that concealment increasingly involves controlling both the signals reaching a sensor and the way intelligent systems interpret them.
In nature, survival depends on more than matching a background. Snow leopard markings disrupt visual recognition, while cephalopods use responsive skin structures to change their appearance. These biological strategies are inspiring engineered materials that adjust colour, reflect light and manage heat. The review highlights a transition from copying outward appearances toward reproducing the coordinated functions behind
natural camouflage.
Modern sensors, however, see far beyond human vision. Infrared cameras detect heat, while radar uses microwaves to reveal objects hidden from sight. Effective
multispectral camouflage must therefore manage several signatures simultaneously. The article describes approaches including adjustable infrared emissions, insulating structures and programmable metasurfaces that control electromagnetic scattering.
Achieving this coordination remains difficult. A material suited to absorbing radar signals may conflict with the requirements for suppressing thermal visibility. Emerging materials, including
graphene and phase change systems, offer possible routes forward, alongside AI tools that help identify material compositions and structures.
Artificial intelligence also introduces another dimension: objects can remain visible yet be misinterpreted by automated vision.
Adversarial attacks exploit weaknesses in recognition algorithms through designed image changes or physical patterns. The review considers these vulnerabilities across digital images, physical objects and systems combining multiple sensing channels, highlighting implications for autonomous driving, robotics and other applications that depend on reliable perception.
Barriers separate laboratory demonstrations from dependable deployment. Manufacturing complex biological architectures at scale remains challenging, while changing illumination, viewing angles and sensor noise can undermine adversarial effects outside controlled conditions. Better understanding these limitations is essential to assessing performance.
Looking ahead, the review outlines opportunities for
adaptive materials, integrated spectral control and algorithms that evolve alongside their countermeasures. Connecting these areas could advance camouflage and stealth while strengthening
robust sensing and
trustworthy AI. As machines gain new ways to see, understanding how perception can be disrupted will become increasingly important to building systems that interpret the world reliably under demanding, changing conditions.
Article Reference
Yongxiang Liu, Mingyang Li, Zhen Liu, Panhe Hu, Jing Tian, Haowen Tan, Muang Ding, Xiaolong Su, Dongfang Guan, Li Liu, The evolution of anti-visual perception: biomimetics, spectral physics, and artificial intelligence,
National Science Review, 2026;, nwag601,
https://doi.org/10.1093/nsr/nwag601
Keywords
anti-visual perception, biomimetic camouflage, multispectral stealth, adversarial attacks, perception manipulation, trustworthy AI