竞争性自适应重加权算法和相关系数法提取特征波长检测番茄叶片真菌病害

发布时间：2018-05-10 11:05

本文选题：高光谱成像技术 + 竞争性自适应重加权算法　；参考：《光谱学与光谱分析》2017年07期

【摘要】：基于竞争性自适应重加权算法(CARS)和相关系数法(CA)特征波长选择方法,提出了利用可见-近红外高光谱成像技术检测番茄叶片灰霉病的方法。首先获取380~1 023nm波段范围内80个染病和80个健康番茄叶片的高光谱图像,然后提取染病和健康叶片感兴趣区域(ROI)的光谱反射率值,作为番茄叶片灰霉病鉴别模型的输入来建立支持向量机(SVM)鉴别模型,训练集和验证集的鉴别率都是100%。研究进一步通过CARS和CA提取特征波长,分别得到5个(554,694,696,738和880nm)和4个(527,555,571和633nm)特征波长,然后分别建立CARS-SVM和CA-SVM鉴别模型。结果显示,CARS-SVM模型中训练集和验证集的鉴别率都是100%,CA-SVM模型中训练集和验证集的鉴别率分别是91.59%和92.45%。以上结果说明了从可见-近红外高光谱图像中提取的光谱反射率值用于检测番茄叶片的灰霉病是可行的。
[Abstract]:Based on competitive adaptive reweighting algorithm (CARSs) and correlation coefficient method (CA), a new method for detecting gray mold in tomato leaves by visible-near infrared hyperspectral imaging (NIR) was proposed. The hyperspectral images of 80 infected and 80 healthy tomato leaves in the range of 380 023nm were obtained, and then the spectral reflectance values of diseased and healthy leaves were extracted. As an input to the identification model of tomato leaf grey mold, support vector machine (SVM) discriminant model is established. The discriminant rate of training set and verification set is 100%. The characteristic wavelengths were extracted by CARS and CA, respectively, and the characteristic wavelengths of 5 (554694696738 and 880nm) and 4 (527555571) and 633nm (respectively) were obtained, and then the identification models of CARS-SVM and CA-SVM were established. The results show that the discriminant rates of training set and verification set in CARS-SVM model are 91.59% and 92.45% respectively. The above results show that the spectral reflectance extracted from the visible-near infrared hyperspectral images is feasible for the detection of grey mold in tomato leaves.
【作者单位】：浙江大学生物系统工程与食品科学学院;浙江经济职业技术学院;
【基金】：教育部高校博士点基金项目(20130101110104);教育部留学回国人员科研启动基金项目国家自然科学基金项目(31471417) 国家(863计划)课题项目(2013AA102301) 中央高校基本科研业务费专项资金项目(2014FZA6005)资助
【分类号】：O657.3;S436.412

【相似文献】