基于多时相哨兵遥感影像的撂荒耕地识别方法

A Method for Identifying Abandoned Farmland Based on Multi-temporal Sentinel-2 Images

利用卫星遥感技术对耕地进行监测具有数据范围大,提取速度快速高效等优点。为此,以湖南省中部某县为例,基于五个时相高分辨率(哨兵2号)卫星遥感影像,利用不同作物时间序列的植被差值指数,对撂荒耕地光谱特征进行分析,采用撂荒耕地的特征值来构建决策树模型,获得各种类型地类空间分布,从而识别出撂荒耕地的分布。研究结果表明,该县撂荒耕地约为20.21万亩,占该县耕地面积25.01%,通过实地样本数据对结果进行验证,结果表明利用卫星遥感影像能够较好地提取出撂荒耕地,正确率高达80.79%。该方法能为我国丘陵山地地区撂荒耕地遥感监测和空间分布提取提供技术参考。

Using satellite remote sensing technology to monitor cultivated land has the advantages of large data range,fast and efficient extraction speed.Therefor,this paper takes a county in Hunan Province as an example,based on five time phases high-resolution satellite(Sentinel-2)remote sensing images,using the difference vegetation index of different crop time series to analyze the spectral characteristics of abandoned farmland,and using the characteristic values of abandoned farmland to build the CART decision tree model to obtain the spatial distribution of various types of land types,so as to identify the distribution of abandoned farmland.The research results show that the abandoned farmland in whole county is about 202.1 thousand mu,accounting for 25.01%of the cultivated land area in whole county.And the results are validated by field samples,the findings suggest that the abandoned farmland can be well extracted by using satellite remote sensing images,and the correct rate is as high as 80.79%.It can provide technical reference for remote sensing monitoring and spatial distribution extraction of abandoned farmland in hilly and mountainous areas of China.