ABSTRACT
Multimodal data, quantified multidimensionally through characteristics obtained from images of plant posture and environmental data, included environmental conditions and growth states affecting the growth and development periods of plants, is expected to gradually change. The Sliding Window-based Support Vector Regression (SW-SVR) allowed development of a model that used a wide variety of complex plant physiological information, realistic measurement data and machine learning with time granularity, to create a soft sensor for wilt detection in plants. The SW-SVR allowed development of a model that used a wide variety of complex plant physiological information, realistic measurement data and machine learning with time granularity, to create a soft sensor for wilt detection in plants. Visual characteristics relating to plant wilting are extracted, using a Convolutional Neural Network, which is one type of deep learning that has been effective in its application to the field of image recognition.
