Classification Of Forest Vegetation Type Using Fused Ndvi Time Series Data Based On Stnlffm

Forest resources play an important ecological role on the earth. The traditional forest resource inventory methods are grueling and time-consuming. It is insufficient by using single-date remote sensing data to identify the forest vegetation types. Fortunately, multi-temporal images have advantages of seasonal variation pattern in different forest vegetation. Duchang County as a study case, GF-1, Landsat8 and MODIS NDVI data were fused to obtain enough spatial and temporal resolution for forest types distinguished by utilizing STNLFFM model and Savitzky-Golay filter. Then, the forest types were classified with the aid of forest inventory data by the SVM method. The classification results showed that the overall accuracy of twice confusion (82%) are significantly better than that of single fusion (72%) only based on Landsat8 and MODIS data. It proved that the more complete temporal data series with higher spatial resolution is very vital even they are derived from different sensors.