Abstract:
Abstract: Remote sensing images with the medium spatial resolution can provide long-time series data of the same area, thus are suitable for remote sensing monitoring of major crops in a large scale. Based on the analysis of the time-series spectrum character curve, crop type identification and acreage extraction can be effectively achieved. The time series curve of normalized difference vegetation index (NDVI) can provide the information of crop growth dynamic change, thus is suitable for remote sensing extracting of major crops planting area. We used Jiangsu province as a research area and employed NDVI (normalized difference vegetation index) time-series data from 46 scenes of MODIS images with spatial resolution of 250 m collected from January 1st 2013 to December 31st 2014, reflectance image data of MODIS collected on April 23rd and image data of Landsat to carry out the remote sensing study for winter wheat planting area. First, a time-series curve of NDVI was built from the MODIS data, which was smoothed by an improved Savitzky-Golay filter. The improved Savitzky-Golay filter reserved the authenticity of data at both ends of the NDVI time series while further improving the smoothness of the curve. Based on the reconstruction of NDVI time series analysis, phenology, plant structure and the samples of ground survey, we extracted the key value of typical objects in their phenological growth period emphatically. At the same time, we analyzed the variation trend of winter wheat, woodland and rice (starting time, range, extent and maximum of NDVI) during the growth period. Through comparing and analyzing the characteristics of NDVI time series curves of different objects after smoothing, we defined the different crops, determined the training rules and build the construction of decision tree so that we can extract the distribution of winter wheat preliminarily. The decision tree classification method can be done quickly and efficiently using multi-threshold, however whose threshold is difficult to select accurately as a result of mixed pixel problem. The range of threshold would affect the accuracy of winter wheat planting area extraction. Therefore, in order to solve the problem of mixed pixels, we used surface reflectance image data to extract the endmember spectral curve of winter wheat. With linear spectral mixture model, according to the abundance ratio of winter wheat, we further extracted winter wheat planting area accurately. The distribution of winter wheat in Jiangsu Province was obtained. At last, the experimental results agreed well with the statistical data and high precision of Landsat8 TM image. According to the results of Landsat TM8 supervised classification, the area of winter wheat was extracted and the results were compared to the frontal results. The precision of decomposing the endmember purified was better than ever. Moreover, it accurately reflected the real situation of winter wheat distribution in Jiangsu province. It was concluded that the method in this paper was simple and easy. Accuracy evaluation results showed that the study area of winter wheat planting area extraction accuracy reached 90%, which can accurately reflect the distribution of winter wheat in the study area, The method also indicated that the application in high resolution remote sensing time-series image data with medium resolution image can be accurately extracted crop acreage, for crop planting area of information extraction to provide reference.