Prediction Of Paddy Rice Crop Variable In Hyperspectral Data Using Backpropagation Neural Network Algorithm

Abstract

Hyperspectral is a new technology in remote sensing which exploits hundreds of bands. Pusat Teknologi Inventarisasi Sumber Daya Alam Badan Pengkajian dan Penerapan Teknologi (PTISDA BPPT) applies hyperspectral in agriculture for yearly yield prediction. In this research, Leaf Area Index (LAI), number of chlorophyll (SPAD), and yearly paddy yield has been predicted with hyperspectral data using backpropagation neural network (BPNN) algorithm. The regions used are Indramayu and Subang; the growth periods of paddy are vegetative, reproductive and ripening, while the heights of the spectral acquisition are 10 cm, 50 cm, and Hymap (2000 m). The data was obtained in cooperation between PTISDA BPPT and ERSDAC Japan. In this study, three test procedures were conducted using backpropagation neural network algorithm, where the first and second procedures used Weka to predict the R2 and RMSE of LAI, SPAD, and yield. Prediction results show high accuracy for LAI but not for SPAD and yield. The third procedure used IDL for prediction yield RMSE, with a satisfactory result : an error value of 0.10153786.