Open AccessOPTIMIZATION OF RICE FIELD CLASSIFICATION MODEL BASED ON THRESHOLD INDEX OF MULTITEMPORAL LANDSAT IMAGES
Author(s) -
Dede Dirgahayu,
Made Parsa,
Sri Harini,
Dony Kurhardono
Publication year - 2020
Publication title -
international journal of remote sensing and earth sciences/international journal of remote sensing and earth sciences
Language(s) - English
Resource type - Journals
eISSN - 2549-516X
pISSN - 0216-6739
DOI - 10.30536/j.ijreses.2020.v17.a3333
Subject(s) - paddy field , enhanced vegetation index , remote sensing , java , index (typography) , scale (ratio) , field (mathematics) , vegetation (pathology) , phenology , vegetation index , computer science , environmental science , leaf area index , statistics , cartography , geography , normalized difference vegetation index , mathematics , agronomy , medicine , archaeology , pathology , biology , world wide web , pure mathematics , programming language
The development of rice land classification models in 2018 has shown that the phenology-based threshold of rice crops from the multi-temporal Landsat image index can be used to classify rice fields relatively well. The weakness of the models was the limitations of the research area, which was confined to the Subang region, West Java, so it is was deemed necessary to conduct further research in other areas. The objective of this study is to obtain optimal parameters of classification model of rice and land based on multi-temporal Landsat image indexes. The study was conducted in several districts of rice production centers in South Sulawesi and West Java (besides Subang). The threshold method was employed for the Landsat Image Enhanced Vegetation Index (EVI). Classification accuracy was calculated in two stages, the first using detailed scale reference information on rice field base, and the second using field data (from a survey). Based on the results of the analysis conducted on several models, the highest accuracy is generated by the three index parameter models (EVI_min, EVI_max, and EVI_range) and adjustable threshold with 94.8% overall accuracy. Therefore this model was acceptable for used for nationally rice fields mapping.