色散模糊解耦的光谱计算成像研究

Spectral Imaging is of great significance in many important areas, e.g., chemical component analysis, material recognition, environmental monitoring and medical diagnosis. However, many shortcomings like the low frame-rate, the low resolution, the insufficient signal to noise ratio, the high system complexity，the calibration errors and the reconstruction difficulties caused by the ignoring of the priors of spectral images, limit the application of the existing spectral imagers. .This project plans to study the spectral image reconstruction from a single dispersive blurred image. The new spectral sensing principle will be investigated at the beginning, and the coupling mechanism of the spectral information contained in the dispersive blurred images will be studied, so that the theory of spectrum sensing from the dispersive blur can be derived. Then, by using the priors of spectral images from the dispersive blur, the spectral image reconstruction technique will be studied. Finally, the spectrum acquisition mechanism based on the dispersive blurred images will be investigated to build the efficient practical acquisition system..The target of this research is to investigate a new spectral imaging method, and provide a new spectral imaging system which is of high acquisition efficiency (large light throughput) and high reconstruction quality (high signal to noise ratio).

光谱成像在诸多重要领域如化学成分分析、材质识别、环境监测、医疗诊断中具有重要的作用。然而，当前光谱成像技术广泛存在分辨率低、系统复杂和精度不足等问题，实际应用受到很大制约。.本项目拟从研究光谱信息在色散模糊图像中的耦合机制以及解耦方法出发，探索建立色散模糊解耦的新型光谱感知理论；进而，研究自然光谱图像数据的先验概率分布规律，探索自然光谱图像先验约束的色散模糊解耦优化方法，突破光谱图像计算重建的关键技术；最后，研究基于色散模糊耦合的新型光谱采集机制，构建高效光学采集系统。.本研究的新型光谱成像方法具有高采集效率(大通光量)和高重建精度(高信噪比)的优点，且系统无需复杂标定。

本项目针对现有光谱成像技术面临的通光量低、信噪比差、光谱重建精度不足，难以应用中/低照度下自然场景（无补光）的动态光谱成像等问题展开研究，提出光谱信息在色散模糊图像中的耦合机制以及解耦方法，建立色散模糊解耦的新型光谱感知理论，构建高效光学采集系统，突破色散模糊耦合的光谱图像计算重建关键技术。项目搭建并实现了具有高采集效率(大通光量)和高重建精度(高信噪比)的色散模糊耦合光谱成像系统，相关研究成果发表在CVPR、ICCV、OE等相关领域的顶级会议和期刊上，共完成9篇文章，9项专利申请，搭建完成实验验证系统，技术指标达到或超过了项目要求。在项目执行过程中，申请人凭借项目相关研究成果获得电子学会2019年中国电子学会技术科学技术奖技术发明一等奖（第三完成人）。.本项目的研究成果有望突破光谱成像技术在通光量和信噪比方面的先天缺陷，推动中/低照度条件下自然场景动态光谱成像技术的实际应用。