The most dangerous disease in countries like India, South Korea and Japan is Apple Marssonina blotch (AMB). In those countries, it causes high loss in economy. Defoliation is caused by AMB. The qualities as well as the quantity of harvested apples are degraded by this disease. Deep neural network is used to rectify this problems in existing techniques. Hyperspectral data is classified with minimized dimension. In this Fully Connected Network (FCN) is follows the Convolutional Neural Network (CNN). It classifies six apple leaf conditions. But noises in the images were not efficiently removed and still the classification result is not accurate. To avoid this issue in this work proposed one efficient noise removal algorithm using wavelet-based (WT) method. Enhancement using image brightness normalization. And in this work feature selection is done by using ant colony optimization algorithm and finally classification of leaf diseases will be computed by using Deep Learning with Support Vector Machines (DL-SVM). Experimental results of proposed classifiers are compared with traditional classification techniques. Qualitative and quantitative analysis carried out on apple leave image collections, demonstrate the robustness of the proposed technique.
Volume 11 | 11-Special Issue
Pages: 223-228
DOI: 10.5373/JARDCS/V11SP11/20192951