基于多光谱遥感的撂荒地识别方法研究

Study on the Method of Recognizing Abandoned Farmlands Based on Multispectral Remote Sensing

耕地撂荒严重影响了我国的粮食产量,成为中国近20年来耕地利用过程中出现的又一重大问题。多光谱遥感具有数据获取范围广、速度快等特点,在国土资源研究中具有非常大的优势和潜力。为了提取撂荒地,文章采用多光谱遥感数据Modis/Terra 2000年—2009年的NDVI数据产品和日本ALOS卫星遥感影像数据,利用时间序列NDVI来表征主要土地利用类型的生命周期特征,对各土地利用类型采样点生成的NDVI时间序列曲线,结合各类型的生长周期特征及其对应的NDVI特征,进行对比分析,从而识别撂荒地,并区分耕地撂荒与轮休。最后通过野外考察开展实地验证,结果表明应用多光谱遥感的方法能够很好的提取撂荒地,并区分主要土地覆盖类型,正确率高达90%。因此多光谱遥感在撂荒地识别研究中具有相当的可行性。

Being abandoned for farmland seriously affected China＇s grain output for farmlands.It has become an important phenomenon over the past 20 years in China.Multispectral remote sensing has the advantage of wide range and high speed in requiring data.It has great potential in the research on land use.Therefore,to extract abandoned farmland in China,the authors＇ used the NDVI data of Modis/Terra from 2000 to 2009 which is one of multispectral remote sensing data and the Remote Sensing Image of ALOS satellite in Japan.The authors＇ used the parameter of NDVI of time series to describe the character of the main land use types.After drawing the time-series curves of the main land use type samples,the authors＇ analyzed them with consulting the life character of these types.Then,the authors＇ compared these curves;finally we recognized abandoned farmland from the others.At last the authors＇ went to experimental plot to survey the land use.The results demonstrated that the method of using multispectral remote sensing data can abstract abandoned farmland and classify the main kind of land use,and the accuracy is as high as 90%.So the method is feasible in recognizing abandoned farmland.