Application of AI Technology to Smart Agriculture: Detection of Plant Diseases

In the agricultural field, crop diseases and physiological disorders have a large impact on its yield and quality. Minimizing damage is very important, but most of the work such as patrols for pest control is done by humans, which is a heavy burden for farmers. Therefore, it is required to reduce the burden by using an automatic detection system for the diseases. In this paper, hyperspectral imaging and AI technology were applied to detect gray mold on tomato leaves. By using probabilistic latent semantic analysis (pLSA) and Bayesian network, optimum 8 wavelengths were selected for machine learning. The prediction accuracy using the selected wavelengths was not significantly decreased compared to using all wavelengths. The results show that pLSA can perform clustering of multidimensional spectral data without compromising the original information, and that Bayesian network is a powerful technique to visualize relationships between multiple variables.