基于近红外漫反射光谱的多品种桃可溶性固形物的无损检测

Nondestructive detection of soluble solids content for multiple peach fruits using near-infrared diffuse reflectance spectra

【目的】研究基于近红外漫反射光谱的多品种桃可溶性固形物含量的无损检测技术。【方法】在获得3个不同品种桃近红外漫反射光谱的基础上,采用多元散射校正(MSC)方法处理原始光谱,以SPXY算法划分样品集,分别建立了可溶性固形物含量的偏最小二乘回归(PLSR)、极限学习机(ELM)和最小二乘支持向量机(LSSVM)预测模型,并比较和评价了移动窗口偏最小二乘法(MWPLS)和连续投影算法(SPA)优选有效特征波长对于简化模型运算量、改善模型预测性能的影响。【结果】虽然全光谱可以获得较好的识别效果,但是模型比较复杂;MWPLS与SPA优选的有效特征波长均能有效地减少建模变量并简化模型,但MWPLS在提高建模效率和改善模型预测精度方面有更明显的优势;PLSR、ELM与LSSVM模型都取得了较理想的预测结果,其中PLSR方法较适用于全光谱建模分析;MWPLS-ELM模型对样品集中桃可溶性固形物含量的预测性能最好,其校正相关系数、校正均方根误差、预测相关系数和预测均方根误差分别为0.991,0.397,0.983和0.497。【结论】近红外漫反射技术可用于多品种桃可溶性固形物含量的准确、无损检测,也为其他品种果品的内部品质指标快速、无损检测提供了技术借鉴。

[Objective] This study aimed to evaluate the prospect of near-infrared （NIR） diffuse reflec- tance spectra in detecting soluble solids content （SSC） of multiple peach fruits. [Method] NIR diffuse re- flectance spectra of peach fruits of three varieties were obtained. The original NIR spectra were prepro- cessed using multiplicative scatter correction method （MSC）, and samp[e sets were partitioned by SPXY al- gorithm. The partial least squares regression （PLSR）,extreme learning machine （ELM） and least squares support vector machine （LSSVM） were established to determine SSC of peach fruits,respectively. The ef- fectiveness of selected characteristic wavelengths in simplifying operation load and improving modeling effi- ciency and prediction performance by moving window partial least squares （MWPLS） and successive pro- jections algorithm （SPA） was compared and evaluated. [Result] Both MWPLS and SPA played an effective role in reducing modeling variables and simplifying models. MWPLS performed better than SPA in enhan- cing modeling efficiency and improving prediction accuracy. PLSR, ELM and LSSVM all had ideal predic- tion results. PLSR could be applied in full spectra model, and MWPLS-ELM gave the optimal prediction performance in detecting SSC of sample sets,with correlation coefficient of calibration of 0. 991,root mean square error of calibration of 0. 397, correlation coefficient of perdition of 0. 983, and root mean square error of prediction of 0. 497. [Conclusion] NIR diffuse reflectance spectrometry could determine SSC of multiple peaches accurately and non-destructively and may be applied to other fruits.