A DEEP LEARNING APPROACH FOR ENHANCING CROP DISEASE DETECTION AND PESTICIDE RECOMMENDATION: Tri-bridNet WITH COLLABORATIVE FILTERING

To ensure the crop health and optimize output in a sustainable way are challenges in the agricultural sector. To meet these challenges, the objective should be prompt detection of crop diseases and the accurate pesticide prescriptions. We present a novel methodology which combines the models of Deep Learning (DL) with a sophisticated image processing method. Both of the metadata and the image data were employed in this work which undergoes to a distinct pre-processing. The segmentation of pre-processed images was used by the model of LeNet-DLV3. By using the statistical features, domain-specific image features, the pertinent features and color features were recovered by the crop image collection as well in metadata. For the Feature Selection (FS), the Tsallis entropy based Conditional Mutual Information (TE-CMI) has been presented. Next, the creation and training of a Tri-bridNet Disease Classifier (TDC) for precise detection of crop disease using Gated Recurrent Units (GRUs), architectures, Convolutional Neural Networks (CNNs) and Multilayer Perceptron (MLP) has been described. After that a strategy of cooperative filtering based on crop disease trends is given along with the environmental variables to recommend the pesticides.