基于便携式仪器的小麦面粉品质检测方法研究

Study on quality testing method of wheat flour based on portable instrument

中文关键词:  偏最小二乘法  遗传算法  预测模型  筛选优化  小麦面粉

英文关键词:partial-least-squares  genetic algorithm  prediction model  screening optimization  wheat flour

中文摘要:

      目的 建立一种噪声小、自适应程度高的小麦粉预测模型的遗传算法-最小二乘法(genetic algorithm partial least squares, GA-PLS)检测小麦面粉品质。方法 采用AMBERⅡ手持式蓝牙光谱仪采集小麦面粉的近红外光谱, 将采集的全波段光谱分段成波长相等的子区间, 对每段子区间的小麦粉水分、灰度以及面筋含量进行最小二乘法预测建模(PLS模型), 将每段建模数据进行遗传算法筛选优化, 最终建立GA-PLS模型, 并与未分段的全谱段PLS模型进行对比分析。结果 基于遗传算法结合偏最小二乘的模型验证精度高于全谱段PLS模型,。其中小麦粉灰度值的相关系数(r2)由0.679上升至0.919, 小麦粉水分的r2由0.701上升至0.923, 小麦粉面筋的r2由0.821上升至0.925。结论 该方法结果准确, 精度高, 适用于小麦面粉品质的现场快速检测。

英文摘要:

      Objective To establish a wheat flour prediction model with low noise and high adaptability for determination the wheat flour quality by genetic algorithm-least squares (GA-PLS). Methods The near-infrared spectrum of wheat flour was collected by AMBERII handheld Bluetooth spectrometer. The collected full-band spectrum was segmented into sub-intervals of equal wavelength, and the least squares prediction modeling of wheat flour moisture, gray scale and gluten content in each sub-interval were carried out. The PLS model was used to optimize the genetic algorithm for each segment of modeling data, and finally the GA-PLS model was established and compared with the unsegmented full-spectrum PLS model. Results The accuracy of the model based on genetic algorithm combined with partial least squares was higher than that of the full-spectrum PLS model. The correlation coefficient (r2) of wheat flour gray value increased from 0.679 to 0.919, the r2 of wheat flour moisture increased from 0.701 to 0.923, and the r2 of wheat flour gluten rose from 0.821 to 0.925. Conclusion The method is accurate and accurate, and suitable for the field rapid detection of wheat flour quality.