摘要
为研究基于国产GF-2影像提取城市绿地的适用方法,文章以一景滁州市城区GF-2影像数据为数据源,分别采用决策树、支持向量机、随机森林、贝叶斯、K最近邻法5种面向对象的分类方法和传统的最大似然法对两个范围大小相同的实验区的绿地进行了提取,并对6种分类方法的分类结果及精度进行了对比评价。结果表明:基于GF-2影像,面向对象分类方法的总体精度较好且相对于最大似然法,总体精度有明显提高;在面向对象分类方法中,随机森林表现效果最好,其总体精度和Kappa系数分别91.17%和0.87,决策树和支持向量机总体精度达到80%以上,K最近邻法和贝叶斯法较差。
In order to study the applicable method of extracting urban green space based on domestic GF-2 image,this paper takes GF-2 image data of Chuzhou City as a data source,uses the Decision Tree(DT),Support Vector Machine(SVM),Random Forests(RF),a Bayesian classifier(Bayes),K Nearest Neighbor(KNN),five object oriented classification methods and traditional Maximum Likelihood(ML)method to extract urban green space from two test areas of the same size,at the same time,we compare and evaluate the classification results and accuracy of the six classification methods.The results show that the overall accuracy of the object oriented classification method is better based on the GF-2 image,and the overall accuracy is significantly improved compared with the Maximum Likelihood method.In the object-oriented classification methods,the Random Forest has the best performance,and its overall accuracy and Kappa coefficient is 91.17%and 0.87 respectively.The overall accuracy of Decision Tree and Support Vector Machine reach over 80%,and K Nearest Neighbor and Bayesian methods are poor.
作者
李伟涛
LI Weitao(College of Geographical Information and Tourist,Chuzhou University,Chuzhou 239000,Anhui,China)
出处
《安顺学院学报》
2019年第3期125-130,135,共7页
Journal of Anshun University
基金
安徽省教育厅自然科学研究重点项目“基于多影像对象特征的森林植被类型信息分层提取方法研究”(项目编号:KJ2017A413)
关键词
GF-2
面向对象
绿地提取
方法比较
GF-2
object oriented
green space extraction
methods comparison
