Optimized recurrent neural network-based early diagnosis of crop pest and diseases in agriculture

The productivity of agriculture plays a critical role in the Indian economy. Growing crop production is a critical responsibility nowadays to accommodate citizen demand and provide farmers with greater rewards. Therefore, a machine learning (ML) technique is employed to more precisely identify diseases and pests on leaves and other crop parts. This paper introduces a machine learning-based system in early crop disease and pest detection using image processing and optimization. Initially, the data is collected from the CCMT plant disease Dataset. Image augmentation techniques such as rotation, flipping, and zooming are utilized to make the dataset wholesome. After amplification, the pre-processing is carried out on these images. Noise reduction as well as enhancing quality are done by Adaptive Bilateral Filter. Lanczos interpolation technique resized it and normalization is done so that the analysis can proceed. Kapur's Entropy-based Whale Optimization is introduced for the segmentation of the image efficiently by dividing diseased areas into segments. The features are extracted using the Gray Level Co-occurrence Matrix, which assesses relationships among the pixels and produces an appropriate feature matrix for color images. This processed data then feeds into a Moth-Flame Optimized Recurrent Neural Network for crop disease and pest detection. These results achieved high accuracy levels at 98.4% for cashews, 98.3% for cassava, 98.5% for maize, and 96.8% for tomato crops, outperforming all the reported techniques.