基于多时相遥感数据的吉林盐碱区土地覆被信息提取方法对比——以镇赉为例

发布时间：2018-03-13 22:42

本文选题：多维分类特征数据集　切入点：支持向量机　出处：《科学技术与工程》2017年05期 　论文类型：期刊论文

【摘要】：利用多时相的遥感数据制作的多维分类特征数据集,可以充分挖掘遥感影像中的植被信息提高地表覆被信息的分类精度。以世界三大盐碱土分布区之一的吉林省镇赉县为例,利用多时相Landsat8遥感数据制作的多维分类特征数据集,通过不同的分类方法提取了实验区11类地表覆被信息,并进行精度对比分析。结果表明:1支持向量机(SVM)法对苏打盐碱化土壤特殊生态环境的地表覆被信息提取具有较好的分类效果,总体分类精度87.77%,Kappa系数0.864 9;其中盐碱地的分类效果较好,生产精度达到98.34%。2不同方案分类精度从高到低依次为:支持向量机、最大似然分类、神经网络、最小距离、光谱角法。3镇赉县的土地利用类型以旱地、水田、盐碱地为主,镇赉西部以旱地为主要,中部地区盐碱地、碱泡、旱地交错分布,东部以水田为主。
[Abstract]:The multi-dimensional classification feature data set made from multi-temporal remote sensing data can fully mine vegetation information in remote sensing images to improve the classification accuracy of surface cover information. Take Zhenlai County, Jilin Province, one of the three major saline-alkali soil distribution areas in the world as an example. Using multi-temporal Landsat8 remote sensing data, the multi-dimensional classification feature data set is used to extract 11 kinds of surface cover information from experimental area by different classification methods. The results show that the SVM method has a good classification effect on the extraction of soil surface cover information from the special ecological environment of soda-salinized soil. The overall classification accuracy is 87.77 and the Kappa coefficient is 0.864 9, in which the classification effect of saline-alkali soil is better, and the production precision reaches 98.34.2 the classification accuracy of different schemes is in order from high to low: support vector machine, maximum likelihood classification, neural network, minimum distance. The main types of land use in Zhenlai County are dryland, paddy field and saline-alkali land, the western part of Zhenlai is dryland, the central area is saline-alkali land, alkali bubble, dryland interlaced distribution, and the eastern part is paddy field.
【作者单位】：吉林大学地球探测科学与技术学院;
【基金】：东北地区国土遥感综合调查中国地质调查局项目(12120115063701)资助
【分类号】：TP79

【相似文献】