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

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。

Detection of Fungal Disease on Tomato Leaves with Competitive Adaptive Reweighted Sampling and Correlation Analysis Methods

Detection of grey mold on tomato leaves using hyperspectral imaging technique based on competitive adaptive reweighted sampling (CARS) and correlation analysis werestudied in this paper .Hyperspectral images of eighty healthy and eighty infected tomato leaves were captured with hyperspectral imaging systemin the spectral region of 380~1023 nm .Spectral reflectanceof region of interest (ROI) from corrected hyperspectral image was extracted with ENVI 4.7 software .The support vector machine (SVM ) model was established based on full spectral wavelengths .It obtained a good result with the discriminated accuracy of 100% in both training and testing sets .Two novel wavelength selection methods named CARS and CA were carried out to select effective wavelengths ,respectively .Five wavelengths (554 ,694 ,696 ,738 and 880 nm) and four wavelengths (527 ,555 ,571and 633 nm) were obtained .Then ,CARS-SVM and CA-SVM models were established based on the new wave-lengths .CARS-SVM modelobtained good results with the discriminated accuracy of 100% in both training and testing sets .CA-SVM modelalso performed well with the discriminated accuracy of 91.59% in the trainingset and 92.45% in thetesting set .It demonstrated that hyperspectral imaging technique can be used for detecton of grey mold disease on tomato leaves .