摘要
为了能够客观、快速、无损、便捷地检测花生仁霉变和出芽,研究了一种基于傅里叶变换近红外光谱技术和K最近邻（KNN）模式识别方法的霉变和出芽花生识别方法。依据花生的感官特征和前人研究经验,将花生分为正常、轻度霉变、重度霉变和发芽四类,采用傅里叶变换近红外光谱仪的积分球漫反射方法采集花生光谱（波段4 000~10 000 cm-1）。利用二阶导数算法进行光谱预处理,建立联合区间偏最小二乘（Si-PLS）识别模型,并得到特征光谱区间。然后在特征光谱区间的基础上运用主成分分析进行数据空间降维,最后建立KNN识别模型。KNN模型训练集与预测集识别率均为98.84%,表明应用近红外光谱技术和KNN法检测霉变和出芽花生效果良好,具有可行性。
A nondestructive testing method was studied for moldy and budding peanuts by Fourier transform near infrared spectroscopy combined with K Nearest Neighbor pattern recognition method. According to the sensory characteristics of peanut and previous research experience,the samples were divided into 4 classes： normal,mild mildew,putrefactive spoilage and budding. Integrating sphere diffuse method was used for the collection of peanut spectra,the wave band was from 4 000 ~ 10 000 cm-1. Second order derivative method was applied for spectroscopy data pretreatment and Synergy Interval Partial Least Squares（ Si-PLS） method was used for the spectral variables selection. The principal components analysis（ PCA） was used for the data space dimensionality reduction. Then KNN recognition model was established. The identification rates of KNN model training set and forecast set were both98.84%. The result indicated that near infrared spectroscopy technology plus KNN method was feasible for nondestructive testing of moldy and budding peanuts with good effect.
出处
《中国农业科技导报》
CAS
CSCD
北大核心
2015年第5期27-32,共6页
Journal of Agricultural Science and Technology
基金
江苏省高校自然科学研究重大项目(14KJA550001)
江苏省第四期"333工程"培养资金项目(BRA2015320)
江苏省高校优势学科建设工程项目资助
