手持式马铃薯干物质含量无损检测装置设计与试验  

Design and experiment of a handheld non-destructive detection device for potato dry matter content

作　　者：丛杰 张悦如 李禧龙 潘宇轩 吕黄珍[2] 吕程序[2] CONG Jie;ZHANG Yueru;LI Xilong;PAN Yuxuan;LYU Huangzhen;LYU Chengxu(School of Mechanical and Electrical Engineering,Qingdao Agricultural University,Qingdao 266109,Shandong,China;National Key Laboratory of Agricultural Equipment Technology,China Academy of Agricultural Mechanization Sciences Group Co.,Ltd.,Beijing 100083,China)

机构地区：[1]青岛农业大学机电工程学院,山东青岛266109 [2]中国农业机械化科学研究院集团有限公司农业装备技术全国重点实验室,北京100083

出　　处：《浙江农业学报》2024年第4期943-951,共9页Acta Agriculturae Zhejiangensis

基　　金：现代农业产业技术体系建设专项资金项目(CARS-10)。

摘　　要：面向马铃薯品质抽检的需求,研发了手持式马铃薯干物质无损检测装置。装置硬件部分包括光谱采集模块、电路控制模块、控制与显示模块、外壳模块,装置设计为枪形,尺寸为180 mm×85 mm×210 mm。利用装置采集可见近红外漫反射光谱,比较Savitzky-Golay卷积平滑(SM)、一阶导数(first-order derivative,FD)、多元散射校正(multiple scattering correction,MSC)和标准正态变量变换(standard normal variate transformation,SNV)的预处理方式,SM结果较优。采用竞争性自适应重加权采样(competitive adapative reweighted sampling,CARS)筛选27个特征波长,建立马铃薯干物质含量的支持向量回归(support vector regression,SVR)预测模型,结果显示,验证集决定系数和均方根误差分别为0.802和0.98%。基于QT开发工具编写装置软件,包括黑白校正与测量模块、电量显示模块、光谱数据显示模块、保存数据模块、光谱数据刷新模块、检测结果显示模块。开展装置验证,预测均方根误差为1.01%,单次测量平均耗时为0.62 s。结果表明,手持式马铃薯干物质无损检测装置可快速、准确检测干物质含量,具有在马铃薯生产源头与加工现场应用的潜力。In response to the demand for quality inspection of potatoes,a handheld non-destructive detection device for potato dry matter was developed.The hardware components of the device included a spectrum acquisition module,a circuit control module,a control and display module,device housing module.The design of the device was gun-shaped with dimensions of 180 mm×85 mm×210 mm.The device utilized visible-near infrared diffuse reflectance spectroscopy,comparing pre-processing methods such as Savitzky-Golay convolution smoothing(SM),first-order derivative(FD),multiple scattering correction(MSC),and standard normal variate transformation(SNV),with SM yielding better results.Competitive adaptive reweighted sampling(CARS)was used to select 27 feature wavelengths to establish a support vector regression(SVR)prediction model for potato dry matter content,with validation set coefficient of determination and root mean square error of 0.802 and 0.98%,respectively.The device software was developed using the QT development tool,including modules for black and white calibration and measurement,power display,spectral data display,data storage,spectral data refresh,and detection result display.Device verification was conducted,showing a prediction root mean square error of 1.01%and an average time consumption of 0.62 s per measurement.The results indicated that the handheld non-destructive detection device for potato dry matter could rapidly and accurately detect dry matter content,demonstrating potential for application at the source and processing sites of potato production.

关 键 词：马铃薯 干物质含量 手持式装置 可见近红外漫反射光谱 

分 类 号：TH79[机械工程—仪器科学与技术]