Classification of Paddy Growth Phase Based on Landsat-8 Image with Convolutional Neural Network Algorithm

Abstract Food security is one of the major issues in the world. One of the related component of the issue is the provision of agricultural data. In Indonesia, there is a difference of rice production data between the Ministry of Agriculture and Statistics Indonesia (locally known as BPS). This is due to difference in data acquisition methods used. In order to predict rice production, we must know about paddy growing phase so that the prediction of rice production in a certain period can be accurately calculated. This research aims to propose a convolutional neural network method to develop paddy growth phase classification model using remote sensing data of landsat-8 images. The convolutional neural network algorithm is used to perform self-learning processes such as reading image, extracting and classifying by changing the structure of landsat-8 image into matrix or pixel. This research also conducted experiment against hyper parameter epoch with 10, 30 and 50 epochs to get good accuracy. The case research is the area of PT Sang Hyang Seri rice field in Subang Regency, Indonesia. The experiment results provided an accuracy of 80.37%. It means the convolutional neural network model is matched with the data. It is able to classify landsat-8 images.