面向对象的覆膜农田信息遥感表征方法

为解决复杂土地利用背景下覆膜农田信息遥感提取方法缺乏的问题,该文以河套灌区为研究区,以Sentinel-2A卫星数据为基础,结合面向对象影像分析和随机森林算法,开展了复杂土地利用背景下灌水与无灌水覆膜农田信息遥感同步提取研究。首先进行遥感影像尺度分割研究,优选出最佳分割尺度。在此基础上,提取光谱特征、纹理特征、几何特征,获取优化特征子集,并采用随机森林机器学习算法表征覆膜农田信息。研究表明,结合利用Sentinel-2A数据与OBIA方法能够有效表征覆膜农田信息,总体精度达93.03%,Kappa系数为0.91;其中,灌水覆膜农田用户精度为91.35%,制图精度为88.57%;无灌水覆膜农田用户精度为97.10%,制图精度为98.63%。研究证明了Sentinel-2A卫星数据和OBIA方法和机器学习算法在覆膜农田信息遥感表征中的应用潜力,能够为地物信息遥感表征研究中提供参考依据。

Characterizing the plastic-mulched farmland using object-based image analysis

To develop a method for characterizing Plastic-Mulched Farmland(PMF)under the complicated land use background,this paper taken Hetao Irrigation District in Inner Mongolia Autonomous region as a research area and based on the Sentinel-2 A satellite data to develop a method for characterizing PMF under the complex land use environment using object-based image analysis(OBIA)and Random Forest(RF)algorithm.First,the images are segmented by using a multi-scale segmentation algorithm,and the optimal segmentation scale is selected.Then,the spectral features,texture features and geometric features were extracted and optimized.Finally,a RF classifier is used to map PMF.The results indicate that Sentinel-2 Aimage and OBIA can effectively extract the PMF information.The overall accuracy can reach 93.03%and the Kappa coefficient is 0.91.the user’s accuracy and the producer’s accuracy of irrigated PMF are 91.35%and 88.57%,respectively.The user’s accuracy and the producer’s accuracy of non-irrigated PMF are97.10%and 98.63%,respectively.This result can prove the potentiality of Sentinel-2 Adata,OBIA and machine learning algorithm for characterizing PMF,and can provide a reference for the research on remote sensing characterization of surface information.