Smart Agriculture Framework for Multi-Crop Disease Detection using IoT and Vision Transformer
Bhanu Sekhar Guttikonda, M Sowmya., Zahraa Alaa Al-Khafaji, P. Senthil, P. Kavitha
Abstract
In recent years, smart agriculture has turn out to be the transformative step to improve crop productivity and sustainability through the integration of Internet of Things (IoT) and Artificial Intelligence (AI), which enabled the continuous monitoring of farms and automated crop health assessment. Although, existing Convolutional Neural Network (CNN) based smart agriculture system faced several challenges such as poor generalization on single crop diseases, unreliable data acquisition in diverse field environments and reduced applicability in remote farming regions. To overcome these challenges, a Vision Transformer (ViT) based smart agriculture system is proposed. This system architecture is composed of four layers, initially, data acquisition layer, which is deployed with sensor nodes and cams for collection of environmental parameters and plant leaf images. Next, the communication layer which is responsible to perform transmission of data to next layer. Here, the next layer is the processing layer where, preprocessing of environmental data is performed using Simple Moving Average (SMA) filtering and threshold evaluation. Further the image data is resized using bilinear interpolation and normalized. Next, this preprocessed data is fed to pretrained ViT model for multi crop and multi disease plant classification. Lastly, the output which includes detected disease types and their confidence levels are sent to the user through an interactive web-based dashboard. Experimental results demonstrates that the proposed ViT based system attained higher accuracy and outperformed the existing system.