基于Radarsat-2的水稻种植面积提取

Paddy rice planting area extraction based on Radarsat-2 data

选用2013年7月23日-10月27日期间5期分辨率为5.2 m×7.6 m的Radarsat-2影像为数据,采用支持向量机法(SVM)和最大似然法(MLC)分别对各时相水稻种植面积进行提取,并以地面实测GPS水稻样方进行精度验证。结果表明SVM和MLC方法的水稻面积提取精度均在9月9日达到最高,所以选择在9月9日的水稻面积提取结果上研究耕地地块优化和碎小图斑去除对精度的影响。通过耕地地块优化和碎小图斑去除处理,水稻面积提取精度显著提高,SVM法由原先的72.876%提高到95.482%,MLC法由74.224%提高到91.792%。

The 5 scenes of Radarsat-2 satellite image with spatial resolution of 5.2 m×7. 6 m collected from July 23rd 2013 to October 27th 2013 were used to extract the paddy rice planting area of every scene using support vector machine （ SVM） and maximum likelihood classification （ MLC） . The accuracy was verified by on-site GPS measurement quadrat ar-eas. Since the extration accuracies of both SVM and MLC were the highest on September 9th, the scene extracted on Sep-tember 9th was chosen to study the effect of farmland parcel optimization and pattern spot removal on the accuracy. The ac-curacy of SVM was improved from 72. 876% to 95. 482%, and the accuracy of MLC was improved from 74. 224% to 91. 792%.