Study On Meatreament Of Wheat Planting Area Using P6 AWiFS Imagery Based On Mixed-Pixel Decomposition Combined With Decision Tree

It is very important to real-timely monitor the wheat planting area for national food security. Remote sensing is a key tool for accurately measuring the area since it is periodical, economical and macroscopic. However, the existing remote sensing method for measuring wheat planting area is hard to put into practice. The reasons are as follows: 1) from the aspect of data resource: medium-high resolution remote sensing data is difficult to access, and the low resolution remote sensing data, such as MODIS, has low measurement accuracy; 2) from the aspect of measuring method: classification accuracy is affected by the phenomenon that same land cover has different spectral, same spectral represents different land covers and mixed pixel; sampling method is lack of geographical information; 3) from the aspect of measuring time: the acquiring time of remote sensing data is always in the period of wheat maturing, it is late for making agriculture policy; 4) from the aspect of adaptability, most methods are just fit for the specific area, they can not be used for large area measuring.Accordingly, a remote sensing method for measuring wheat planting area is presented in this paper, which will be robust and can be used in practical operation in large regions. Two IRS-P6 AWiFS imageries, flight at Oct. 22nd and Nov. 119th covering Beijing City, was chosen according to wheat phonology calendar. Firstly, 292 field points was sampled based on analysis the remote sensing data to get land cover types. The field points'NDVI was analyzed to get the characteristic of wheat's NDVI. Then a decision tree was designed to derive wheat pixel. Finally, wheat pixel was unmixed based on NDVI using linear mixture modeling to get a higher accuracy.Applying the method to Tongzhou and Daxing, which are of different planting structures, the result shows that wheat distribution's region accuracy and pixel accuracy is up to 95% and 85% respectively. The conclusions are drawn as follows:1) IRS-P6 AWiFS is a good data source to derive data of wheat, which can effectively support the measurement of wheat planting area;2) It is effective to measure wheat planting area with remote sensing method at the beginning period of wheat grow in North China.3) The integration of mixed-pixel decomposition technology and decision tree improves the accuracy of area measurement.4) This method is highly robust, better than the most commonly used methods.