AUGMENTING SMART FARMING: SEVERITY-CENTRIC BLAST PADDY LEAVES ASSESSMENT
Abstract
Paddy leaf rice blast disease is a devastating rice disease which causes severe yield losses and threatening rice production globally. Rice blast caused by the fungus Magnaporthe oryzae, pose a threat to both the above and below-ground parts of paddy plants. It's crucial to identify signs of disease, like rice blast severity and understand its effective management strategies. Deep learning (DL) methods have shown to be successful in solving this problem for intricate prediction problems. The study presents a unique Spatial Stacked Deep Convolutional Neural Network (SS-DCNN) method for forecasting the impact of a rice blast severity that utilizes DL principles. Using a prominent disease dataset from Kaggle, it was implemented a three-step preprocessing methodology, including k-means based segmentation, high-pass filtering, and bicubic interpolation, to ensure data quality and enhance images. To extract meaningful features, it employs two methods, Histogram of Oriented Gradients (HOG) and Accelerated-KAZE (AKAZE), which aid in reducing dimensionality while retaining essential features. The proposed model is applied to finish the classification task, naturally accounting for temporal correlations in the data. This is especially important for estimating the intensity of the rice blast severity, as past trends have a big impact on results. The study, which focuses on rice blast severity forecasting, is carried out with the help of Python tools. Several measures of accuracy are used to evaluate the suggested model's efficiency. Their all-encompassing strategy seeks to improve rice blast severity prediction measures relative to current approaches.