苹果叶部病害的傅里叶变换红外光谱鉴别研究

Identification of Apple Leaf Disease Type Based on Fourier Transform Infrared Spectroscopy

为建立一种基于傅里叶变换红外光谱技术结合光谱检索和逐步判别分析的苹果叶部病害快速鉴别方法,以白粉病、花叶病、炭疽叶枯病和早期落叶病4种病害,共60份样本的红外光谱、一阶导数光谱和二阶导数光谱为指标,利用Omnic 8.5软件中光谱检索功能依次与相应光谱库进行检索鉴别。检索结果显示:基于一阶导数光谱和二阶导数光谱的检索正确率均为96.7%,高于基于红外光谱检索的83.3%。同时,以样品光谱差异较大的1 800~1 000 cm^(-1)波数内的二阶导数红外光谱数据作为判别变量,利用SPSS 20.0软件中的逐步判别分析功能,比较了基于5种挑选判别变量方法建立的判别模型的鉴别效果。鉴别结果显示:采用Mahalanobis距离逐步判别法建立的模型对苹果叶部病害的鉴别效果最好,对训练样本的回判正确率为100.0%,对测试样本的预测正确率为80.8%,总正确率最高,为92.3%。综上表明,傅里叶变换红外光谱技术结合光谱检索法或逐步判别分析法,均能较好地诊断苹果叶部病害种类,可为苹果叶部病害的鉴别和诊断提供一种省时、易行的方法。

Fourier transform infrared （ FTIR） spectroscopy combined with spectra retrieval and stepwise discrimination analysis was used to identify the disease type of apple. The infrared spectra of 60 samples from 4 diseases （ powdery mildew, mosaic disease, anthrax blight and early deciduous disease） were collected and the first-derivative spectra and second-derivative spectra for all samples were calculated by the software Omnic 8. 5, and three spectra libraries were constructed, separately. The absolute differential difference search using the infrared spectra, first-derivative spectra and second-derivative spectra was carried out, respectively. Results showed correct rate of 96. 7% for the first-derivative spectra and second- derivative spectra, and 83. 3% for the infrared spectra. The second-derivative spectra in the range of 1 800-1 000 cm 1 were used to build models by stepwise discrimination analysis. The discrimination effectiveness was compared for the five ways to choose discrimination variance, and results showed that the model based on Mahalanobis distance method was better for the identification of apple disease type. The correct rate of returned classification reached 100. 0% ,the predication accuracy was 80. 8% ,and the total correct rate was 92. 3% . The results indicate that as a time-saving and convenient method, the FTIR spectroscopy combined with spectra retrieval or stepwise discrimination analysis is feasible to distinguish the diseases of apple.