IoT Enabled Efficient Detection and Classification of Foliar Disease in Apple Trees

Abstract

Controlling the outbreaks of pests and diseases in agricultural environment is still an enormous challenge to the farmers. Apples, one of the most important and widely cultivated fruit crops in the world, are also constantly under the threat of being affected by a large number of insects, pathogens and are susceptible to a wide range of diseases. This in turn affects the overall productivity and more often than not downgrades the quality of the fruit. Such diseases are currently diagnosed in orchards by manual scouting, which takes intensive manual labour and is expensive in terms of money and time. Timely detection of diseases can save these plants from severe damage and reduce the disease rate. In this research, wepropose a method for multi-label foliar (leaf) disease classification using deep learning models incorporating it with an IoT farmland monitoring system for rapid and automated responses.We have implementedtwo deep learning models– InceptionV3 and Xception. Xception was found to slightly outperform InceptionV3 and the F1 scores achieved were 95.2 and 94.1 respectively.