用沃罗诺伊图生成结构化对抗伪装，高效且难察觉。
Structured Adversarial Camouflage via Voronoi Diagrams
- 通过优化种子点位置生成有结构的对抗图案，无需额外正则化。
- 在3DPeople数据集上应用时，检测准确率大幅下降（AP@[.5:.95]显著降低）。
- 对颜色调换敏感，适合关注视觉隐蔽性和对抗鲁棒性的研究者。
像素级对抗补丁计算成本高且易被察觉，限制其在安全关键系统中的应用。本文提出基于沃罗诺伊图的对抗伪装方法，仅优化种子点位置，使用固定可打印色板与软分配机制，生成类似碎裂状的结构化对抗图案，无需额外正则化。在行人检测任务中，基于原始图像的简单放置（Inria → COCO）表现较差，而通过分割掩码实现衣物级应用（3DPeople）导致检测精度显著下降。攻击在跨域背景和不同检测器（YOLOv9/10/11/12）间有效迁移，表明其在黑盒场景下的鲁棒性。更换色板几乎完全消除攻击效果，单色微调容忍度极低（≤0.17），凸显结构与配色间的强耦合关系。该参数高效、色板受限的设计提升了视觉合理性，同时损害实时检测性能。物理验证与色彩校准留待未来工作。代码见：https://github.com/JensBayer/Voronoi。本论文原发表于2026年5月12-13日英国巴斯举行的国际军事通信与信息系统会议（ICMCIS），由信息科技技术委员会（IST）主办，编号IST-224-RSY。
原文摘要 · Abstract (English)
Pixel-wise adversarial patches are computationally heavy and often visually detectable, limiting utility in security-critical systems. We present adversarial Voronoi camouflage that optimizes only seed-point locations under fixed, printable palettes using a soft assignment, producing structured, splinter camouflage-like patterns without additional regularization. Evaluated on person detection with COCO-style AP@[.5:.95], naive placement (Inria -> COCO) performs comparably bad, while garment-level application via segmentation mask (3DPeople) results in a significant AP drop. The attack transfers to out-of-domain backgrounds and across detector families (YOLOv9/10/11/12), indicating robustness in black-box settings. Repainting with different palettes largely nullifies the effect, and single-color tweaks show limited tolerance (<=0.17), highlighting a structure-palette coupling. The parameter-efficient, palette-constrained design improves visual plausibility while degrading real-time detector performance. Physical validation and color calibration are left for future work. Code: https://github.com/JensBayer/Voronoi This paper was originally presented at the International Conference on Military Communication and Information Systems (ICMCIS), organized by the Information Systems Technology (IST) Scientific and Technical Committee, IST-224-RSY - the ICMCIS, held in Bath, United Kingdom, 12-13 May 2026.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。