PSO Based Deep Belief Networks Learning for IoT based Crop Disease Detection on Paddy Leaves Using Cloud

The Internet of Things (IoT) with advanced machine learning techniques presents significant potential for agricultural applications, particularly in the domain of crop disease detection. Paddy, a staple food for millions, is highly susceptible to various diseases that can drastically affect yield and quality. Early and accurate disease detection is crucial for effective management and mitigation. Traditional methods are often labor-intensive and less reliable, underscoring the need for automated, accurate, and scalable solutions. The primary challenge lies in developing a robust system capable of accurately identifying diseases in paddy leaves using IoT-collected data. This task is complicated by the variability in disease manifestation and environmental conditions, which can affect the quality and consistency of the collected data. Efficient feature extraction and classification techniques are essential to address these issues and ensure high accuracy. This study proposes a novel approach combining Particle Swarm Optimization (PSO) for feature extraction with Deep Belief Networks (DBNs) for classification. IoT devices capture highresolution images of paddy leaves, which are then processed in the cloud. PSO is employed to optimize the feature extraction process by selecting the most relevant features from the image data. These optimized features are fed into a DBN, which is trained to classify the images into healthy or diseased categories. The use of cloud computing ensures the scalability and computational efficiency of the system. The proposed method demonstrates significant improvements in accuracy and processing speed. The PSO-based feature extraction enhances the relevance of features, reducing the dimensionality and improving the DBN's performance. Experimental results show an accuracy rate of 96.3%, with a reduction in processing time by 35% compared to traditional methods. The system's precision and recall rates are 95.8% and 94.7%, respectively, highlighting its effectiveness in real-world applications.