Crop Disease due to viral, fungal and bacterial types impacts the yield potential of the crop and economic damage to the farmers.EspeciallyAnthracnose is group of fungal diseases typically cause dark lesions on leaves. It is difficult to phytosanitary control and will easily spread through the air and abundant presence in various climates especially tropical and continental which causes the loss of numerous crops and thus leads to large economic damage to farmers. Classification of crop species may lead to misinterpretations. It is to develop an efficient Spectral Angle Mapper system for the detection of anthracnose and other fungal diseases in the cropsusing hyperspectral images. For this purpose, Discriminative Marginalized Convolution Neural Network has been employed as classification methods on spectral data. The control measures can be improved by mapping the occurrence of spectral signatures of fungal disease specific end members. In this paper, Crop Disease Classification is carried out with respect to fungal disease on various climatic conditions. Initially image preprocessing is carried out to eliminate the spectral noises. Preprocessed image is processedusing independent Component analysis to extract the endmembers of the spectral band representing the various crop species in the different climate regions like tropical and continental. Extracted spectral band is processed for spatial correction and the extreme spectral variability detection through spectral purity index. Spectral member of various spatial regions is selected using genetic algorithm. Finally Discriminative Marginalized Convolution Neural Network is employed to generate the Spectral angle mapper to classify the spectral band of the specific crop species in various climates on aspect of Chlorophyll and Nitrogen spectral values of the end members. Proposed classifier uses the multiple layer of network to discriminate the crop diseases related to fungal disease as healthy, moderately infected, and severely infected plants on 14 crop species on. Finally utilization of the disease index represents the severity of crop.The experimental results of the proposed model are evaluated on employing real-time hyperspectral image data sets acquired from Indiana region in the North Karnataka on various climates. Performance of the proposed model is compared with conventional approaches with respect to precision, recall and f measure.

Hyperspectral Image, Convolution Neural Network, Fungal Disease, Crop Classification, GeneticAlgorithm, Independent Component Analysis, End Member, Vegetation Index