傅里叶变换红外光谱数据超分辨率处理关键问题研究

Fourier transform infrared spectra often suffer from common resolution degradation problems of spectral noise, baseline drift and bands overlap in the process of data acquisition. In this project, to overcome those challenges, we carried out the research on infrared spectrum super-resolution. Combined with the physical properties of the Fourier transform spectrometer, we try to build the degradation model of infrared spectrum data and research on the generation mechanism of band overlap and spectral noise. Then, the nonlinear solution algorithms for the parametric instrument response function (IRF) model will also be studied in our research. Furthermore, guided by this model, we plan to establish a unified super-resolution processing framework for the Fourier transform spectrum data. We study the spectral detail-preservation methods, which is combined with the spectral characteristics, to suppress the spectral strong noise. Based on the wavelet transform of the spectroscopic data, we will study the adaptive spectral restoration methods under the condition of the low signal-to-noise ratio. Those methods will integrate the spectral local structure information. At last, we plan to research on spectral quality and performance evaluation of the super-resolution algorithm. We plan to do the multiple factors and levels orthogonal experiments to determine influence law between the parameters change and the super-resolution algorithm performance. After those experiments, the optimal combination of the infrared spectrometer parameters will be put forward, which can improve the quality of spectrum and super-resolution processing performance. The research will provide new methods, new technology and evaluation methods for the Fourier transform infrared spectrometer and its super-resolution processing. And also, the research will promote the practical application of the infrared spectrum analysis in the field of target recognition, material identification and chemometrics analysis, etc.

针对傅里叶变换红外光谱在数据获取过程中存在的光谱噪声、基线漂移和谱带重叠等分辨率退化问题进行超分辨率处理研究。结合傅里叶变换光谱仪的物理特性，建立光谱数据退化降质模型，重点研究谱带重叠和噪声产生的机理、仪器响应函数参数化模型非线性求解问题。在模型的指导下，建立傅里叶变换光谱数据超分辨率处理统一框架，研究适应光谱自身特性的细节保存方法抑制光谱强噪声，低信噪比条件下基于小波变换融入光谱局部结构信息的自适应光谱复原方法。开展光谱质量与超分辨率算法性能评价研究，采用多因素多水平正交实验法确定工作参数变化与超分辨率算法性能的影响规律，从提高光谱质量与超分辨率处理性能角度提出工作参数的最优组合。研究成果将为傅里叶变换红外光谱仪及其超分辨率处理提供新方法、新技术和评价方法，促进红外光谱分析在目标识别、物质鉴定、化学计量学分析等领域的实际应用。

本项目按照资助项目计划书，针对傅里叶变换红外(Fourier transform infrared spectrum, FTIR)光谱数据在获取过程中存在的随机噪声、谱带重叠和基线漂移导致的谱线分辨率降低问题，开展红外光谱分辨率提高的理论模型、方法算法和仿真实验研究，主要研究成果如下：.（1）建立反映随机噪声、谱线重叠和基线漂移的FTIR光谱数据退化数学模型，将光谱数据超分辨率增强转化为反问题求解问题，建立了基于最大后验概率估计理论的FTIR光谱数据噪声抑制、基线校正和光谱复原处理统一框架。.（2）深入挖掘红外光谱中漂移基线的本质属性，提出一种基于自适应Tikhonov正则化约束的联合噪声抑制和基线校正方法。该方法能有效地校正任意漂移的基线和随机噪声，同时保存谱线结构特征；.（3）融入FTIR结构信息和小波频域变换稀疏性先验知识，提出融入谱线结构自相似性特征的字典学习光谱复原方法，以及小波变换系数稀疏性先验的红外光谱数据复原方法。提出的算法能有效地增强光谱分辨率并移除光谱噪声。.（4）将课题红外光谱超分辨率方法应用到了Raman光谱、毫米波图像的分辨率增强方面，有效地提高了Raman光谱的谱分辨率和毫米波图像的清晰度。.课题组已经完成本课题的研究内容，达到了预期的研究目标，并进一步推广应用了本课题的相关成果。本课题发表了标注本基金号资助学术期刊论文10篇，学术会议论文4篇，申请专利4项。培养毕业硕士生3名，在读博士生3名，硕士生5名。执行期间课题主持人入选2016年度“香江学者”人才计划，赴香港城市大学交流2年，并于2018年6月晋升副教授。在基金资助下，课题组已成长为包含研究生在内的10余人的研究团队，在红外光谱数据复原、噪声抑制和数学模型优化方法等方面取得了一定结果和相关积累。