基于高光谱成像技术和机器学习的猕猴桃果实可溶性固形物含量预测  

Prediction of soluble solids contents in kiwifruit based on both hyperspec-tral imaging technology and machine learning

作　　者：刘子涵 李明[3] 赵峙尧 陈谦 李佳利 于家斌[1] 钱建平 LIU Zihan;LI Ming;ZHAO Zhiyao;CHEN Qian;LI Jiali;YU Jiabin;QIAN Jianping(School of Computer and Artificial Intelligence,Beijing Technology And Business University,Beijing 100048,China;Institute of Agri-cultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences/State Key Laboratory of Efficient Utilization of Ar-id and Semi-arid Arable Land in Northern China,Beijing 100081,China;Zhengzhou Fruit Research Institute,Chinese Academy of Agri-cultural Sciences/National Key Laboratory for Germplasm Innovation&Utilization of Horticultural Crops,Zhengzhou 450009,Henan,China)

机构地区：[1]北京工商大学计算机与人工智能学院,北京100048 [2]中国农业科学院农业资源与农业区划研究所·北方干旱半干旱耕地高效利用全国重点实验室,北京100081 [3]中国农业科学院郑州果树研究所·果蔬园艺作物种质创新与利用全国重点实验室,郑州450009

出　　处：《果树学报》2024年第12期2606-2620,共15页Journal of Fruit Science

基　　金：国家重点研发项目(2022YFF1101103、2022YFD1600703、2022YFD1600700);黑龙江省“揭榜挂帅”科技攻关课题(2021ZXJ05A0501);中国农业科学院科技创新工程(CAAS-ASTIP-2023-ZFRI-03);北京市自然科学基金项目(4222042,6242004);河南省重点研发专项(221111111800);国家现代农业产业技术体系(CARS-26);河南省现代农业产业技术体系(HARS-22-09-S)。

摘　　要：【目的】可溶性固形物含量(SSC)是评价猕猴桃果实品质的关键指标。旨在利用高光谱技术构建猕猴桃果实SSC预测方案,实现无损、准确评估果实内部品质。【方法】以米良一号猕猴桃果实为研究对象,对高光谱图像进行白板校正、感兴趣区域提取;采用MSC、SG平滑、SG-MSC和SG-SNV方法进行光谱数据预处理以消除噪声影响,并通过PLSR模型确定最优方法;结合CARS、SPA和RF算法分别提取与果实SSC相关的特征波段;建立PLSR、SVR、RFR、BPNN模型,比较特征波段与SSC实测值之间的耦合关系,选出最优模型,并利用PSO算法优化其预测精度,以实现果实内部品质的泛化预测。【结果】MSC方法在全波段回归中表现最佳;CARS算法有效简化模型并提取关键特征波段;SVR模型预测精度最高,经PSO优化后训练集和测试集决定系数分别为R_(c)^(2)=0.949,R_(P)^(2)=0.913;均方根误差分别为RMSEC=0.341 2,RMSEP=0.364 9。【结论】相比于单一环节的算法优化,MSC+CARS+PSO-SVR的组合模型在猕猴桃果实可溶性固形物含量预测方面表现更优,研究结果可为果品品质监测和分级分选提供技术支持。【Objective】In the context of predicting soluble solids contents(SSC)for Miliang No.1 ki-wifruit,SSC is a key quality indicator representing the concentration of soluble sugars,which are impor-tant for determining the sweetness and maturity of the fruit.Accurate and timely SSC assessment is cru-cial for both consumer satisfaction and market pricing.Traditional methods like refractometry and liq-uid chromatography,while accurate,are time-consuming,costly and destructive,making them unsuit-able for large-scale or real-time monitoring.To address these challenges,this study aims to develop a non-destructive SSC prediction model using hyperspectral imaging technology,integrating multiple pre-processing methods,feature extraction algorithms and machine learning models.The goal is to enhance the robustness and generalization of SSC predictions by optimizing the entire prediction process,rather than focusing on individual steps like preprocessing or feature extraction,which has been the primary focus of many previous studies.【Methods】This study was conducted using 150 Miliang No.1 kiwi-fruit samples,which were randomly divided into a training set of 120 samples and a test set of 30 sam-ples.Hyperspectral images were captured using a Rikola portable hyperspectral imager,covering the spectral range from 500 nm to 900 nm with a wavelength interval of 2 nm,resulting in 194 spectral bands.The imaging was conducted in a controlled dark-box laboratory environment to ensure data con-sistency and minimize external interference.After the hyperspectral images were captured,SSC mea-surements were performed using an ATAGO PAL-BX/ACID 8 refractometer.Three SSC measurements were taken for each sample,and the arithmetic mean of the three values was used as the actual SSC val-ue.To improve the quality of the spectral data,various preprocessing methods were applied.Four spe-cific methods were employed to enhance data consistency and eliminate noise:multiplicative scatter correction(MSC),Savitzky-Golay smoothing(SG),SG combined with

关 键 词：猕猴桃 高光谱成像技术 可溶性固形物含量 机器学习 品质预测 

分 类 号：S663.4[农业科学—果树学]