Enhanced Sugarcane Leaf Disease Detection via Hybrid Deep Capsule Autoencoder CNN Optimized by Improved Whale Algorithm

This research uses a Hybrid Deep Capsule Autoencoder-based Convolutional Neural Network and an Improved Whale Optimization (IWO) model to classify sugarcane leaf diseases. It starts with public dataset images that are pre-processed using an adaptive intensity-limiting histogram equalization technique. This preprocessing technique improves data quality for analysis by increasing input image contrast and drastically reducing noise. Hybrid wavelet-Walsh transform and Laplacian edge detection with a Gaussian operator capture leaf image properties during feature extraction. Adaptive Horse Herd Optimization (AHO) decreases dimensionality while keeping crucial information for accurate classification to improve the feature set. The suggested Hybrid model employs capsule networks and convolutional neural networks to classify sugarcane leaf diseases, with the IWO model improving performance. Testing shows that the framework works, with 97.213% accuracy. This finding outperforms previous methods and shows hybrid deep learning strategies' promise in agricultural disease identification. In precision agriculture, the findings provide a powerful tool for early sugarcane leaf disease detection and control, improving crop output and sustainability.