Estimation Of Rice Phenology Data And Yield Using Long-term NDVI Data From AVHRR

With the development of remote sensing, traditional agriculture is also continuously developing to high-tech agriculture, using remote sensing technology as a means to provide effective technical support for the development of agricultural information technology. Not only large area of crop area extraction, development time extraction, long-term potential monitoring, production estimation is possible, but also provide a large number of image data and analysis methods for many other problems.The application of long time series vegetation index data sets is more and more concerned because of its long time span and its ability to reflect the seasons and interannual changes of crops in a long period of time. In this paper, using AVHRR NDVI time series data to extract the crop growth period, first of all solve the problem of the original time series data denoising, use several filtering denoising method of original time series data, get a smooth time series curve, and then use different filtering method to reconstruct the time series data set to extract the key still growing rice.Using GIMMS NDVI time series data for rice yield estimation and to trend method in separating tendency yield and output of remote sensing, and then build a remote sensing output and yield estimation factor regression model, and finally to rice yield estimation remote sensing in Jiangsu province. The results showed that filtered long-time AVHRR NDVI data could extract the transplanting, maturing and heading date of rice and control the error within ± 16 days. The identification error of heading date and maturity was controlled at Within 6 days. The estimated yield model established by de-trended yield can predict the yield of rice in Jiangsu Province. The relative error between predicted yield and statistical yield is within 10%.