Automatic Identification of Plant Physiological Disorders in Plant Factories with Artificial Light Using Convolutional Neural Networks

Abstract

Plant factories with artificial light (PFAL) are attracting worldwide attention as a technology for stably producing crops. One of the problems of PFAL is tipburn which is a physiological disorder of crops. Tipburn is a phenomenon in which plant growth point cells are necrotized. Lettuce cultivated in PFAL in particular has a high frequency of tipburn. When tipburn occurs, its identification is done by human eye observation, and tipburn leaves are trimmed by hand or tipburn lettuce is removed from products. These operations require much labor and cost. If tipburn identification can automatically be done using machine learning, the economic effect will be great and it will be a driving force for spreading PFAL. In this study, we aim to perform binary discrimination of tipburn occurrence and its non-occurrence about lettuce cultivated in PFAL using machine learning with convolutional neural networks. In particular, we aim to recognize the symptom of tipburn which means the early stages of tipburn immediately before leaf tips discolor blackly and the commercial value as the vegetables is damaged. The results of the experiments indicate that the recognition of the symptom of tipburn can be performed with high accuracy.