基于近红外光谱的果树残枝纤维组分含量分析

【目的】研究运用近红外光谱技术结合化学计量学实现快速检测新疆南疆果树残枝中纤维素、半纤维素和木质素含量。【方法】以150个从新疆南疆各地采集的果树残枝样本为材料,利用近红外光谱技术结合偏最小二乘法（PLS）,采用不同的预处理和特征波段筛选方法优化各纤维组分含量的预测模型。【结果】SG卷积平滑法预处理结合竞争性自适应权重取样法（CARS）优选特征波段建立的3种纤维组分近红外检测模型效果最优,相关系数r分别为0.950 3、0.948 7和0.937 1,决定系数R²分别为0.900 8、0.896 5和0.875 1,校正标准偏差RMSEC分别为0.007 0、0.005 4和0.005 1,预测标准偏差RMSEP分别为0.011 8、0.008 9和0.008 8。【结论】采用近红外光谱技术能够实现新疆南疆果树残枝纤维素、半纤维素和木质素三组分的快速定量检测。

Determination of Fiber Component Content in the Residual Branches of Fruit Trees in South Xinjiang Based on Near Infrared Spectroscopy

【Objective】 The detection of fiber components in fruit tree stumps generally has the problems of time-consuming, complicated operation and high test cost. The research uses near-infrared spectroscopy technology combined with chemometrics to quickly detect the content of cellulose, hemicellulose and lignin in the stumps of fruit trees in southern Xinjiang. 【Method】 Taking 150 samples of fruit tree stumps collected from various parts of southern Xinjiang as the research object, using near-infrared spectroscopy technology combined with partial least squares（PLS）, Using different pre-processing and characteristic waveband screening methods to optimize the prediction model of the content of each fiber component. 【Result】 The three fiber component near-infrared detection models established by the SG convolution smoothing method preprocessing combined with the competitive adaptive weight sampling method（CARS） optimized feature band have the best effect, and the correlation coefficients r are 0.950,3, 0.948,7 and 0.937,1, respectively. The coefficients of determination R² are 0.900,8, 0.896,5, and 0.875,1, the corrected standard deviations RMSEC were 0.007,0, 0.005,4, and 0.005,1, and the predicted standard deviations RMSEP were 0.011,8, 0.008,9, and 0.008,8, respectively. 【Conclusion】 The use of near-infrared spectroscopy technology can achieve rapid quantitative detection of the three components of cellulose, hemicellulose and lignin in fruit tree stumps in South Xinjiang.