Multi-modal sliding window-based support vector regression for predicting plant water stress

Abstract

Information communication technology (ICT) is required in the field of agriculture to solve problems arising because of the aging of farmers and shortage of heirs. In particular, environmental sensors and cameras are widely used in existing agricultural support systems for easy data collection. Although the traditional purpose of these systems is naive monitoring and controlling of the environment, the propagation of advanced cultivation is now expected by applying the data to machine learning and data mining technologies. Therefore, we propose a novel multi-modal sliding window-based support vector regression (multi-modal SW-SVR) method for accurate prediction of complicated water stress, which is a plant status, from two data types, namely environmental and plant image data. The proposed method includes two methodologies, SW-SVR and deep neural network (DNN) as a multi-modal feature extractor for SW-SVR. SW-SVR, which we proposed previously, is a suitable learning method for data with time-dependent characteristics, such as plant status. Moreover, we propose a new image feature, remarkable moving objects detected by adjacent optical flow (ROAF), to enable DNN to extract essential features easily for predicting water stress. Compared with existing regression models and features, the proposed multi-modal SW-SVR with ROAF demonstrates more precise and stable water stress prediction.