Rice Extraction Based On Fused Data Of GF-1 And MODIS In POYANG Lake Area

Food is the most basic means for a human to exist,China is a large agricultural country which has a large population,the demand pressure on rice especially on food is also increasing,so,to obtain the crop distribution fast,timely and accurately has an important influence on the social economy,food security,ecological function and so on.We can use remote sensing technology to estimate crop area in large scale,but the existing remote sensing data can't both meet requirements of high time resolution and high spatial resolution,which greatly limits the monitoring,STARFM can generate high spatial resolution data which can provide new ideas on the extraction of crop planting area.This paper took Poyang Lake District of Jiangxi Province as the research object,according to the MOD09Q1 and GF-1/WFV data in 2014 to conduct high spatial resolutio n remote sensing data set based on STARFM model,use the decision tree classificat io n method to extract large scale crop planting area based on this data set,the main results are as follows:(1)The correlation coefficient of high temporal reflectance fusion data based on STARFM model and real image basically can reach more than 0.6,but with the increase of the time,the growth status of crops changed greatly,the real image and fusion image have a certain difference in the spatial details,the fusion accuracy decreased.(2)Using S-G can smooth the noise generated by cloud and haze of time series NDVI images,so the seasonal variation of the processed images can better reflect the features of the plants.(3)The overall classification accuracy reached more than 83% based on the high spatial resolution fusion remote sensing images,and with the help of crop phenological data,recognition accuracy is higher than time series NDVI image,so the fusion data can solve the temporal limitation problem of the high spatial resolution remote sensing images on crop area extracting.(4)The fusion time series NDVI image with high spatial and temporal resolution can effectively classify the broken objects and have higher recognition accuracy,the recognition effect is better than that of single scene GF-1/WFV and MODIS NDVI data.