New Approach to Crop Disease Classification and Data Security in Smart Agriculture Networks

New Approach to Crop Disease Classification and Data Security in Smart Agriculture Networks

연구 분야: Infrastructure

학회: Cognitive Computation

Agriculture is the backbone of the economy which plays a significant role in the economic life cycle. The traditional secure authentication-based methods are affected by challenging tasks such as low reliability, security deployment, and high monitoring costs. To overcome these challenges, the Smart and Secured Agriculture Crop Management Network model is proposed in this research article. This proposed architecture is created based on different layers such as crop, edge, blockchain, network, and application layers. The crop images are collected from the agricultural farm in the crop layer and the edge layer is used to collect the crop images for further processes. These collected crop images are preprocessed to enhance image quality by implementing imputation, normalization, and data-cleaning processes. The features present in these preprocessed images are extracted by using the MobileNetV2 model that enhances computation complexity. The Kohonen Learning Dense Attention-based transformer model is proposed to categorize the different kinds of crop diseases. The blockchain is applied for securing stored data and the smart contract is introduced in this blockchain layer for storing data in an Interplanetary File System that handles high data storage costs. 6G network interface is used as a network layer and the application layer accelerates scientific discovery and crop productivity by minimizing environmental impacts. To validate the system, classification-based analysis was conducted to assess disease detection accuracy, latency, and processing delays, while security-based analysis evaluated the robustness of blockchain integration, including metrics such as security level and node communication time. The experimental evaluation of the proposed model attained performance values of 98.86%, 0.1 ms, 28 s, 95.9%, and 51 s from accuracy, average delay, latency, security level, and node communication time respectively. These simulation findings show that our proposed framework attained superior results compared to previous works in smart and secured crop management systems.