Publications

Suwantong, R; Srestasathiern, P; Lawawirojwong, S; Rakwatin, P (2016). Accurate Crop Cultivation Date Estimation from MODIS using NDVI Phases and the Extended Kalman Filter. 2016 13TH INTERNATIONAL CONFERENCE ON ELECTRICAL ENGINEERING/ELECTRONICS, COMPUTER, TELECOMMUNICATIONS AND INFORMATION TECHNOLOGY (ECTI-CON).

Abstract
Crop yield forecasting is important either for a government, agricultural industries or a trading company for their action plans. A very important variable to be given to a crop model used for the forecast is an accurate crop cultivation date. When the area of interest is large, it is preferable to use remote sensing data such as satellite images for the crop monitoring. The estimation of the cultivation date can be done using the Normalized Difference Vegetation Index (NDVI) collected from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Terra satellite which is interesting for a developing country thanks to its free and frequent availability. However, in a tropical area the NDVI data is very noisy due to clouds. To cope with this issue, it is proposed in the literature that one can model the NDVI by a triply modulated cosine function with the mean, the amplitude and the initial phase as state variables and then use the Extended Kalman Filter (EKF) to estimate these three variables. However, the EKF tuning and cultivation date computation methods in the literature cannot be used in reality for accurate age estimation. This paper therefore proposes to use the estimated total phase of the triply modulated cosine function to compute crop cultivation date. To specify, the crop cultivation date is defined as the day in which the estimated total phase is higher than a threshold. The threshold is predefined using available real cultivation dates from available ground truths during the same year in the region. Thanks to the method proposed in this paper, the mean cultivation date error is reduced from 11.70 days to 0.0 day and the mean absolute error is reduced from 26.2 days to 16.3 days compared to when the method in the literature is implemented for our study cases which are single crop rice fields in the northeast of Thailand.

DOI:

ISSN: