基于MESMA和RF的山丘区土地利用信息分类提取

发布时间：2018-01-12 01:22

本文关键词：基于MESMA和RF的山丘区土地利用信息分类提取　出处：《农业机械学报》2017年07期 　论文类型：期刊论文

【摘要】：探讨了基于多端元混合像元分解(Multiple endmember spectral mixture analysis,MESMA)和随机森林(Random forest,RF)相结合的土地利用信息分类提取方法。以Landsat-8 OLI卫星遥感影像为主要数据,基于植被-不透水面-裸土(Vegetation-impervious surface-soil,VIS)模型,利用MESMA将影像分解为植被、不透水面和裸土3类组分,将生成的3类组分变量和基于光谱、纹理信息计算选取的20个特征变量组合后开展RF分类实验,将分类结果与相同特征变量下的支持向量机(Support vector machine,SVM)、最大似然(Maximum likelihood classification,MLC)分类结果进行比较分析。结果表明:MESMA可以获得较为精确的组分丰度信息;RF分类结果优于相同特征变量下的SVM和MLC分类结果;在MESMA生成的组分信息变量参与分类后,3种方法的分类精度均有所改善,分别达90.50%、88.85%、86.35%,其中RF的分类精度改善最为显著;MESMA与线性混合分解(Linear spectral mixture analysis,LSMA)生成的组分信息变量相比,前者对于改善分类精度效果更为明显。MESMA对于提高影像分类精度起到一定积极作用,基于MESMA和RF的方法对中等空间分辨率影像山丘区土地利用信息分类提取精度较高,利用该方法开展遥感影像解译可为大尺度的土地利用监测和管理工作提供技术支持和理论参考。
[Abstract]:To investigate the decomposition of mixed pixels based on multi element (Multiple endmember spectral mixture analysis, MESMA) and random forest (Random forest, RF) of land use classification information extraction method combining with Landsat-8. OLI satellite remote sensing data, vegetation impervious surface - based on bare soil (Vegetation-impervious surface-soil, VIS) model, using MESMA image is decomposed into vegetation, impervious surface and bare soil of 3 kinds of components, will generate 3 group variables and based on spectral classification experiment to carry out RF, select the calculation of texture information of 20 variables combined with the classification results of support vector machine with the same characteristic variables (Support vector machine, SVM the maximum likelihood (Maximum), likelihood classification, MLC) classification results were compared and analyzed. The results show that MESMA can obtain more accurate component abundance information; RF classification results are better than the same Characteristic variables under the SVM and MLC classification results; information of variables involved in classification in MESMA generation group, the classification accuracy of the 3 methods were improved, up to 90.50%, 86.35%, 88.85% respectively, which improve the classification accuracy of RF was most significant; MESMA and linear mixture (Linear spectral mixture analysis, LSMA) generation the components of information variables to improve the classification accuracy compared to the former effect is more obvious to.MESMA certain positive role to improve the image classification accuracy, and the method of MESMA based on RF images of medium spatial resolution land use classification information extraction in hilly areas with higher precision, using the method of remote sensing image interpretation can provide technical support and theory the reference for large scale land use monitoring and management.

【作者单位】：中国地质大学(北京)土地科学技术学院;国土资源部土地整治中心;
【基金】：国土资源部公益性行业科研专项(201511010-02)
【分类号】：F301.2;TP751
【正文快照】： 引言准确掌握国家土地资源利用信息是国家相关部门调整土地利用结构、合理开发土地资源、动态监测土地利用状况等工作的基础[1]。土地资源与遥感数据在时空特性方面具有高度一致性,对土地资源的研究成为遥感技术应用的主要领域之一。过去数十年,国内外开展了大量研究,促使了遥

【相似文献】