Applying a Convolutional Neural Network in an IoT Robotic System for Plant Disease Diagnosis

Abstract

Plant diseases are major threat to green product quality and agricultural productivity. Agronomists and farmers often encounter great difficulties in early detection of plant diseases and controlling their potential production damages. Thus, it is of great importance for stakeholders to diagnose plant diseases at very early stages of plant growing by exploiting state-of-the art technologies, consider appropriate actions and avoid further economic losses. Artificial Intelligence (AI) techniques, field sensors, data analytics and inference algorithms are some contemporary tools which could be helpful for early plant disease diagnosis. In this paper, we present a plant Disease Diagnosis Support System (DDSS) that utilizes an Internet of Things platform to control a lightweight robotic system. The DDSS applies a Convolution Neural Network learning algorithm to perform early plant disease diagnosis and classification. The system can help farmers to apply appropriate precision agriculture actions and better control their production. The proposed DDSS achieves around 98% success classification rate, according to our demonstration case study.