基于高光谱技术的玉米种子可视化鉴别研究  

Study on Visual Identification of Corn Seeds Based on Hyperspectral Imaging Technology

作　　者：吴翔[1] 张卫正[1] 陆江锋[1] 裘正军[1] 何勇[1] 

机构地区：[1]浙江大学生物系统工程与食品科学学院,浙江杭州310058

出　　处：《光谱学与光谱分析》2016年第2期511-514,共4页Spectroscopy and Spectral Analysis

基　　金：国家科技支撑计划项目(2014BAD04B04);浙江省教育厅科研项目(Y201328805)资助

摘　　要：种子纯度是衡量种子品质的重要指标。提出一种基于近红外（874~1 734nm）高光谱技术实现玉米种子可视化鉴别的方法。采集4个品种共384个玉米种子样本的高光谱图像数据,随机选择288个样本作为建模集,剩余96个样本作为预测集。对玉米种子光谱曲线进行分析后,通过连续投影算法（SPA）选取7个特征波段作为输入,结合偏最小二乘法判别分析（PLS-DA）模型,对预测集进行预测,获得较好的分类效果,其中R_C=0.917 7,RMSECV=0.444 2;R_（CV）=0.911 5,RMSECV=0.459 9,建模集和预测集的总体鉴别率分别为78.5%和70.8%。通过图像处理技术提取高光谱图像中每个玉米颗粒的平均光谱数据,输入建立的SPA-PLS-DA模型,在计算生成的鉴别图中以不同颜色标识不同类别,实现了混杂玉米种子样本的可视化鉴别。对3份不同组成的混杂种子样本进行鉴别,达到了较好的可视化效果。结果表明,通过可视化鉴别技术,可以直观方便地观察混杂种子样本中不同品种种子的分布和数量,为农业生产中种子的纯度鉴别和筛选提供了帮助。The seed purity is an important indicator of seed quality.The paper proposes a visual identification method of corn seed based on the near-infrared（874~1 734nm）hyperspectral image technology.Hyperspectral image data of 4cultivars of a total of 384 corn seed samples will be acquired.Then 288 of samples are to be selected randomly as the calibration set,and the remaining 96 samples will be used for the prediction set.After inspection of the near-infrared spectral curves,7effective wavelengths（EWs）are to be selected by successive projection algorithm（SPA）.And then 7EWs of the calibration set will be used as input to build a partial least squares（PLS）model.Good results are to be obtained with R_C=0.917 7,RMSECV=0.444 2;R_（CV）=0.911 5,RMSECV=0.459 9.And the total identification rate of the developed PLS model will be 78.5%for the calibration set and 70.8%for the prediction set.Finally,average spectral data of each corn seed in a hyperspectral image will be extracted by image process technology,and used as input of the developed SPA-PLS model.In the produced identification map,different colors are to be used to represent different predicted cultivars.3mixture samples of corn seeds will be identified,and help to achieve satisfied visual effects.The result indicates that,by means of the visual identification technology we could intuitively observe the distribution of corn seeds of different cultivars in mixture samples.The research provides help for the identification and screening of seeds in agricultural production.

关 键 词：高光谱成像 玉米种子 连续投影算法 可视化 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]