REAL-TIME PREDICTION OF CROP DISEASES USING IOT-ENABLED DATA ACQUISITION AND MACHINE LEARNING

Abstract

Crop diseases are a major threat to global agriculture by affecting yield quality and quantity while leading to economic impacts and food insecurity. Risk management measures must be initiated with few steps for early detection and immediate treatment or control in order to mitigate the potential consequences of crop disease. This paper presents a unified proposal for the first real-time prediction of crop diseases using Internet of Things (IoT)-enabled data acquisition fused with machine learning (ML). The proposal is based on the system and includes a collection of IoT devices including environmental sensors, smart-cameras, drones, low-power wide-area networks (LPWANs) for transmitting the data that collect and relay environmental information (e.g., temperature, humidity, soil moisture, plant wellbeing, and other information). Furthermore, all field-oriented data use, relatively speaking, better, centralized cloud storage (and processing) which ensures the data collected from IoT data transmissions is accessible. Preprocessing techniques that cleaned and normalized the data were utilized to enhance the value of the inputs before uploading a data stream. Data would then be fed into a Convolutional Neural Network (CNN) to capture high-resolution images of the leaves in order to assess disease symptoms, compared to other environmental and agronomic data, respectively. Environmental/ agronomic data would provide access to supervised learning models (e.g., Random Forest and Gradient Boosting) for assessing or detecting possible early-stage disease patterns.