Volume - 7 | Issue - 2 | june 2025
Published
20 May, 2025
This study presents an innovative classification framework for detecting diseases affecting mango blossoms and stems by integrating deep learning techniques with a meta-heuristic optimization strategy. A dataset consisting of 3,500 images, collected directly from mango orchards across the Konkan region in Maharashtra, is used for training and evaluation. To enhance image clarity and enable effective feature extraction, Contrast-Limited Adaptive Histogram Equalization (CLAHE) is applied during preprocessing. Deep features are subsequently extracted using Convolutional Neural Networks (CNNs). To optimize the extracted feature space, a novel hybrid algorithm, Adaptive Squirrel-Grey Wolf Search Optimization (AS-GWSO) is introduced, boosting both processing efficiency and classification precision. For the final classification stage, an Enhanced Long Short-Term Memory (E-LSTM) model is employed, which utilizes the optimized features to improve generalization and minimize overfitting. Experimental results demonstrate that the proposed AS-GWSO-LSTM model outperforms existing classifiers and meta-heuristic-based models, achieving a high accuracy rate of 97.2%. The findings highlight the model’s strong applicability for real-time agricultural disease monitoring systems.
KeywordsMango Blossom and Stem Disease CLAHE CNN AS-GWSO Enhanced LSTM Deep Learning Agricultural Disease Detection
