Study On Estimation Methods Of Maize Area Based On GF-1 Data In Changji City

Changji city was selected as the study area.With the intention of the crop classification of high-resolution remote sensing images(GF-1)and maize area evaluation,after masking images and calculating NDVI,NDVI time-series images and five high resolution and wide-swath remote sensing images(GF-1)of 2015 were used to recognize maize individually by maximum likelihood,mahalanobis distance,minimum distance,parallelepiped,neural net classification method and support vector machine.By linear mixed pixel decomposition to improving the interpretation accuracy.Meanwhile,the classification results were compared with the Landsat-8 image classification results.As the results showed that:Firstly,the total extraction accuracy of maize area based on GF-1 image is better than Landsat-8 images,which indicated that the image data of high space and high time resolution was better in recognizing maize.The study area covered by mask could shield the impact of non-cultivated land in the process of the supervised classification,so as to improve the accuracy of interpretation.Secondly,the classification results of GF-1 and Landsat-8 image showed that,whether using single-phase image or NDVI continuous sequence image,the optimal method of crop classification and maize area extraction in Changji city was the maximum likelihood method.Thirdly,using a single GF-1image for maize classification and identification,the higher classification accuracy of the month is July,August,and the total accuracy of maize area could reach to 96.71%,98.47%.Further,using the NDVI time series image,the accuracy was 97.93%.The images of GF-1 in July and August were analyzed respectively by linear mixed pixel decomposition model,the total area accuracy reached to 98.76% and 98.99%.Through a comprehensive comparison,the highest accuracy of maize area was extract from August image analyzed by linear mixed pixel decomposition,and then it was classed by the maximum likelihood method.Finally,the overall accuracy of the crop classification in Changji is 98.61% in 2015,and the total accuracy of maize area is 98.99%.