基于高光谱技术检测小麦粉灰分含量  

Detection of ash content in wheat flour based on hyperspectral technology

作　　者：李佳琮 谷晏 刘翠玲[1,2] 孙晓荣[1,2] 张善哲 LI Jia-Cong;GU Yan;LIU Cui-Ling;SUN Xiao-Rong;ZHANG Shan-Zhe(School of Artificial Intelligence,Beijing Technology and Business University,Beijing 100048,China;Beijing Key Laboratory of Big Data Technology for Food Safety,Beijing Technology and Business University,Beijing 100048,China;Market Supervision Administration of Beijing Xicheng District,Beijing 100035,China)

机构地区：[1]北京工商大学人工智能学院,北京100048 [2]北京工商大学,北京市食品安全大数据技术重点实验室,北京100048 [3]北京市西城区市场监督管理局,北京100035

出　　处：《食品安全质量检测学报》2023年第5期60-67,共8页Journal of Food Safety and Quality

基　　金：北京市自然科学基金项目(4222043)。

摘　　要：目的 基于高光谱技术实现对小麦粉灰分含量的准确检测。方法 利用高光谱成像技术采集小麦粉的光谱数据,建立基于偏最小二乘法(partial least squares regression,PLSR)和深度极限学习机(deep extreme learning machines,DELM)的小麦粉灰分含量预测模型;通过分析3种预处理算法和4种波长选择算法,分别选出最佳的预处理与波长选择方法,最后构建基于特征波段光谱信息的预测模型,并对结果进行比较。结果 标准正态变量校正(standard normal variable,SNV)为最佳预处理方法;连续投影算法(successive projections algorithm,SPA)相较于随机森林(random forest,RF)、无信息变量消除(uninformative variable elimination,UVE)和遗传算法(genetic algorithm,GA)选择特征波长的模型更优;DELM模型更适用于灰分含量的检测,最优模型的测试集决定系数为0.968,预测集均方根误差为0.024。结论 高光谱成像技术可以快速、精准的无损检测小麦粉灰分含量,该技术可为在线检测小麦粉品质系统的开发提供理论依据。Objective To realize the accurate detection the ash content of wheat flour based on hyperspectral technique. Methods A model for predicting the ash content of wheat flour based on partial least squares regression(PLSR) and deep extreme learning machines(DELM) was established by collecting spectral data from wheat flour based on hyperspectral imaging techniques. The best pre-processing and wavelength selection methods were selected respectively by analyzing 3 kinds of pre-processing algorithms and 4 kinds of wavelength selection algorithms.Ultimately, a prediction model based on the spectral information of the characteristic bands was constructed and the results were compared. Results Standard normal variable(SNV) was the best pre-treatment method;successive projections algorithm(SPA) outperformed random forest(RF), uninformative variable elimination(UVE) and genetic algorithm(GA) were better to select the model of characteristic wavelengths;the DELM model was more suitable for the detection of ash content, the test set coefficient of determination of the optimal model reached 0.968, and the root mean square error of prediction reached 0.024. Conclusion Hyperspectral imaging technology allows fast and accurate non-destructive detection of ash content in wheat flour, this technology can provide a theoretical basis for the development of an online system for testing wheat flour quality.

关 键 词：小麦粉灰分含量 高光谱成像技术 深度极限学习机 波段筛选 

分 类 号：O657.3[理学—分析化学] TS211.7[理学—化学]