تشخيص زود هنگام علائم ويروس موزائيك خيار با روش طيف سنجي مرئي- فروسرخ نزديك

چكيده انگليسي :

Abstract According to the increased demand for high quality fruits and vegetables and thedisappearing the trade limitations in recent years the mechanized agriculture has increasinglybeen required Meanwhile early detection of plant diseases which reduces the quality andquantity of products has been widely considered On the other hand cucumber is the firstproduced greenhouse crop in Iran The presence of the disease in this plant causes reducingthe growth and decreasing the yield Cucumber mosaic virus CMV is one of the mostcommon viral diseases with wide host range Early rapid and automatic diagnosis of thisdisease can be effective in controlling loos management and increasing productivity Healthmonitoring and diagnosis of CMV are commonly performed using molecular methods suchas polymerase chain reaction PCR that requires the precise sampling and time consumingexperiments Therefore development of a non destructive method which can early detect thesymptoms of disease seems to be necessary In this study the ability of visible and near infrared Vis NIR spectroscopy to detect the early symptoms of CMV was evaluated Forthis purpose the leaves with the CMV suspected symptoms were first collected byinspecting the cucumber greenhouses and farms around the city of Isfahan In the next step 214 cucumber plants were cultivated 124 of which were infected with CMV by mechanicalinsemination method and the remaining 90 plants were kept intact A photo diode arrayVis NIR spectrometer equipped with a CCD detector with the resolution of 2 nm and therange of 200 1100 nm was used to record the leaf spectra The spectrometer was equippedwith a bifurcated fiber optic a sample holder and a halogen light source After themechanical inoculation of the virus the samples were irrigated regularly for a maximum of17 days until the symptoms emerged During this period spectral collection were carried outevery two days until the end of the seventeenth day Then the samples were divided intothree classes including intact infected with invisible symptoms and infected with visiblesymptoms A total of 3660 spectra were obtained from aforementioned three classes whichwere used for classification analysis In order to analyze the obtained spectra the beginningand the end of the spectra were eliminated due to unwanted noises and then variouspreprocessing methods were applied to the spectra To classify soft and independentmodeling of class analogy SIMCA linear discriminant analysis LDA quadraticdiscriminant analysis QDA support vector machine SVM and principal componentsanalysis combined with artificial neural network PCA ANN were used The results showedthat the best preprocessing method in the SIMCA and LDA were normalizing the spectrawith the total accuracies of 78 28 and 86 44 respectively in test set validation In theQDA the best results were obtained using MSC preprocessing with a test set accuracy of84 45 The SVM method with smoothing preprocessing resulted in a total accuracy of76 61 Finally in PCA ANN method the intact invisible and visible symptoms classeswere distinguished with a total accuracy of 84 37 By comparing different classifiers thebest results were obtained using the LDA method in which the intact invisible and visiblesymptoms classes were discriminated with the accuracies of 100 83 29 and 88 38 respectively and the total accuracy of 86 44 In discrimination the intact and infectedclasses the total accuracy was 100 in fact all intact and infected samples were correctlyclassified On the whole the results of this study showed that the non destructive Vis NIRspectroscopy had a strong potential for detecting the cucumber plants infected to CMV Keywords NIR Spectroscopy cucumber Cucumber mosaic virus Soft independentmodeling of class analogy Linear discriminant Quadratic discriminant Support vectormachine Artificial neural networks 67

صفاري، مهسا

تشخيص زود هنگام علائم ويروس موزائيك خيار با روش طيف سنجي مرئي- فروسرخ نزديك

طيف سنجي NIR , خيار , ويروس موزاييك خيار , مدل سازي نرم و مستقل شباهت هاي بين كلاسي , تفكيك كننده ي خطي , تفكيك كننده ي درجه دوم , ماشين بردار پشتيبان , شبكه هاي عصبي مصنوعي