【摘要】 针对传统近红外光谱仪器体积庞大且成本高昂的现状，基于滨松MEMS-FPI近红外光谱传感器C14272，成功研发了工作波长为1350-1650 nm的微小型化近红外光谱仪。该系统采用模块化架构，整合高效信号获取与数据处理单元，显著降低系统成本。通过系统性能评估，包括波长精度、光谱分辨率、动态范围等指标，并通过温度特性补偿算法有效抑制了温度漂移。实验结果表明：该光谱分析系统具备优异的稳定性与重现性，光谱分辨率约为15.2 nm。结合卷积神经网络算法，构建的葡萄糖溶液浓度预测模型在检测未知样本时表现出高准确性（R~2>0.99）和低误差率（RMSE<0.2%）。所研制的微小型化近红外光谱仪为生物医学检测与食品质量分析提供了经济高效的解决方案。更多还原

【Abstract】 Traditional near-infrared spectrometers are bulky and expensive, limiting their widespread application. Based on Hamamatsu’s MEMS-FPI near-infrared spectroscopy sensor C14272, this research successfully developed a miniaturized near-infrared spectrometer operating in the 1350-1650 nm wavelength range. The system adopts a modular architecture, integrating efficient signal acquisition and data processing units, significantly reducing system costs. Through systematic performance evaluation, including wavelength accuracy, spectral resolution, dynamic range, and other metrics, combined with a temperature characteristic compensation algorithm that effectively suppresses temperature drift effects, experimental results demonstrate that this spectroscopic analysis system possesses excellent stability and reproducibility, with a spectral resolution of approximately 15. 2 nm. Combined with a convolutional neural network algorithm, the glucose solution concentration prediction model exhibits high accuracy(R~2>0. 99) and low error rates(RMSE<0. 2%) when detecting unknown samples. The developed miniaturized near-infrared spectrometer provides an economical and efficient solution for biomedical detection and food quality analysis.更多还原