基于人工蜂群算法优化的SVM遥感分类方法——以玛纳斯湖古湖盆为例  

Study of SVM Classification Method Optimized by Artificial Bee Colony Algorithm:A Case Study of the Ancient Manasi Lake Basin

作　　者：杨雪峰[1] 吐热尼古丽.阿木提 YANG Xue-feng;Tureniguli·AMUTI(College of Geographic Science and Tourism,Xinjiang Normal University,Urumqi 830054,China)

机构地区：[1]新疆师范大学地理科学与旅游学院,新疆乌鲁木齐830054

出　　处：《地理与地理信息科学》2018年第4期40-45,共6页Geography and Geo-Information Science

基　　金：国家自然科学基金项目(41461023);博士科研启动基金项目(XJNUBS1526)

摘　　要：针对如何通过优化支持向量机参数实现更优分类效果的问题,进行了基于人工蜂群算法的支持向量机参数优化研究。首先,基于玛纳斯湖古湖盆区域的多角度遥感数据和土地覆被信息构建观测数据集。然后,分别使用人工蜂群算法和网格法优化的支持向量机、决策树和最大似然法,依据训练集建立分类模型,继而对测试集进行分类。通过对分类混淆矩阵的比较分析发现,在使用该观测数据集的情况下,当以径向基为核函数时,无论利用人工蜂群算法还是网格法对惩罚因子C和核参数γ的取值进行优化,支持向量机的分类效果均优于决策树和最大似然法。另外,经过人工蜂群算法优化后的支持向量机比网格法可以获得更好的参数分布,实现更高的分类正确率。最后,使用人工蜂群算法优化参数后的支持向量机对玛纳斯湖古湖盆区域的土地覆被进行了制图。To solve the problem of optimizing the support vector machine(SVM)parameters to achieve better classification effects,the optimization of SVM parameters based on artificial bee colony(ABC)algorithm was studied.Firstly,based on the multi-angle remote sensing data and land cover information in the ancient Manasi Lake Basin,the observation data set was built.Then,the ABC algorithm and the Grid Search method were used to optimize the SVM parameters;the classification results derived from the ABC_SVM algorithm and the Grid Search_SVM algorithm were compared with the results derived from the J48 decision tree and maximum likelihood classification(MLC)methods.It is found that when using RBF as kernel function and using ABC algorithm or the Grid Search method to optimize the value of the penalty factor C and the kernel parametersγ,the classification effect of the SVM is better than that of the decision tree and the MLC method.In addition,the SVM optimized by ABC algorithm can get better parameter distribution than the Grid Search method and achieve higher classification accuracy.Finally,the land cover in the ancient Manasi Lake Basin was mapped by using the ABC_SVM algorithm.

关 键 词：人工蜂群算法 支持向量机 参数优化 多角度遥感 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]