人眼房水葡萄糖浓度的近红外光谱分析技术研究

发布时间：2018-02-28 07:13

本文关键词： 近红外光谱预测模型波段筛选偏最小二乘法遗传算法　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：近红外光谱分析技术是当今血糖无创检测的一个研究热点。由于近红外光谱分析技术具有非接触、反应灵敏、不需要试纸等特点,对糖尿病的治疗具有重要的实际意义。但是通常被测样品的近红外光谱吸收较弱,谱峰重叠严重,因此选择包含样本信息的建模波段是近红外光谱分析技术需要解决的核心问题。本文主要针对人眼葡萄糖浓度的近红外光谱预测模型进行了研究。首先以葡萄糖溶液的近红外吸收光谱为测试对象,采用传统的偏最小二乘法建立基本的数学校正模型,用均方根预测误差RMSEP和相关系数R作为评价模型的指标;针对传统预测模型稳定性较差和预测精度较低的问题,本文提出了区间组合移动窗口偏最小二乘法和改进遗传算法,筛选出葡萄糖溶液的最优建模波段为1567nm-1641nm;实验配置了含有葡萄糖、尿素、抗坏血酸盐和乳酸盐的模拟房水溶液,采集其吸光度光谱并懫用传统偏最小二乘法和本文提出的两种方法分别建立数学校正模型,并对模型预测精度进行分析。最后基于本文提出的算法,用Matlab软件编写了近红外光谱的数据处理系统,该系统为科研人员提供了一个便捷的光谱分析平台。
[Abstract]:Near-infrared spectroscopy (NIR) is a hot spot in the field of noninvasive blood glucose detection. Because NIR is non-contact, sensitive, and does not require test paper, etc. It is of great practical significance for the treatment of diabetes mellitus. However, the NIR spectra of the samples are usually weak, and the spectral peaks overlap seriously. Therefore, the selection of the modeling band containing sample information is the core problem to be solved in the near infrared spectrum analysis technology. This paper mainly focuses on the near infrared spectrum prediction model of glucose concentration in the human eye. Firstly, we study the near infrared spectrum prediction model of human eye glucose concentration. The near infrared absorption spectra of the solution were measured. The traditional partial least square method is used to establish the basic mathematical correction model, and the root-mean-square prediction error (RMSEP) and correlation coefficient R are used as the indicators of the evaluation model. In this paper, an interval combined moving window partial least-squares method and an improved genetic algorithm are proposed. The optimal modeling band of glucose solution is 1567nm-1641nm.The aqueous solution of simulated room containing glucose, urea, ascorbic acid salt and lactate is configured experimentally. The absorbance spectra were collected and the mathematical correction models were established by the traditional partial least square method and the two methods proposed in this paper, and the prediction accuracy of the model was analyzed. Finally, based on the algorithm proposed in this paper, A data processing system for near infrared spectrum is developed by using Matlab software. The system provides a convenient platform for researchers to analyze the spectrum.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O657.33;R587.1

【相似文献】