高光谱成像
计算机科学
全光谱成像
计算机视觉
光谱分辨率
人工智能
遥感
传输(电信)
滤波器(信号处理)
光谱成像
图像分辨率
计算复杂性理论
可扩展性
忠诚
迭代重建
成像光谱学
窄带
光谱形状分析
光学
高保真
扫描仪
钥匙(锁)
光学滤波器
光谱带
图像质量
波长
计算模型
多光谱图像
作者
Shiqi FENG,Xiao Liang,Dongbai Xue,王绪泉 Wang Xuquan,Siyu Dong,Zhiyuan Ma,Xiong Dun,Zhanshan Wang
出处
期刊:
[SPIE]
日期:2026-10-21
卷期号:6 (01): 1-18
标识
DOI:10.1117/1.apn.6.1.016004
摘要
Short-wave infrared (SWIR) spectral information is critical for scientific analysis and industrial technology. On-chip computational hyperspectral imaging systems with miniaturized footprints and single-shot acquisition have attracted widespread attention. Metasurface-based computational hyperspectral cameras have emerged as a promising implementation; however, research remains comparatively underexplored. One key challenge is pixel-level alignment errors during integration, which severely degrade the quality of reconstructed spectra. In this work, we develop a single-shot on-chip SWIR computational hyperspectral camera based on a metasurface array. To mitigate inevitable pixel-level misalignment, we propose a misalignment-robust design strategy by minimizing spectral transmission differences between neighboring filters and identifying an optimal spatial layout. Simulation results show that our method effectively suppresses degradation in reconstruction accuracy caused by alignment deviations up to 0.5 μm. Experimental results demonstrate that the prototype achieves a spectral resolution of ∼10 nm and an average reconstruction fidelity of 96.81% for narrowband spectra across the 900 to 1700 nm wavelength range. In addition, our prototype delivers high-fidelity spectral reconstruction for representative target scenes and exhibits excellent performance in real-world plastic-sorting tasks. This work establishes a new pathway toward fully integrated, single-shot SWIR computational hyperspectral cameras based on a metasurface scheme, providing a compact, high-speed, and scalable solution for resource-constrained environments.
科研通智能强力驱动
Strongly Powered by AbleSci AI