Snapshot thermal spectral imaging based on hybrid metasurfaces with high efficiency and high spectral resolution

Engineering notes

Key topics: autonomous driving. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Long-wave infrared (LWIR) spectral imaging plays an important role in fields such as machine vision, industrial monitoring, and military reconnaissance. However, current thermal spectral imaging devices are often bulky and complex due to their reliance on intricate optical paths and scanning mechanisms, and their filtering systems face inherent limitations and trade-offs in key performance metrics such as resolution, bandwidth, and light throughput. In this study, we propose an LWIR spectral imaging chip that directly integrates hybrid metasurface filter arrays into an LWIR sensor, enabling chip-scale thermal spectral imaging across 8 to 12 μm. The coupling strength between metasurface resonant modes is enhanced using a particle swarm optimization algorithm. The average optical transmittance of the encoder reaches 67.3%, achieving broadband, high-throughput encoding. With a deep learning network, we achieve a reconstruction speed of 16.1 frames per second (fps) at 1024×768 resolution, a peak signal-to-noise ratio above 40 dB, and a spectral resolution of 1.2 nm (Δλ/λ<0.001). This work demonstrates a compact on-chip platform for real-time, high-spectral-resolution, and high-efficiency thermal spectral imaging, holding promise for applications such as autonomous driving and satellite remote sensing.

Links and sources

Comments