第39卷第2期2023年6月测绘标准化StandardizationofSurveyingandMappingVol.39No.2Jun.2023收稿日期:2022-09-18第一作者简介:凡建林,工程师,现主要从事工程测量、地理信息系统工程等工作。鲸鱼算法优化的SVM区域高程异常拟合方法凡建林1高叶1姚辉2(1.浙江省测绘科学技术研究院浙江杭州310000;2.宁波冶金勘察设计研究股份有限公司浙江宁波315016)摘要:针对采用支持向量机(SupportVectorMachine,SVM)区域高程异常拟合法最佳参数难以确定问题,本文提出采用鲸鱼算法(WhaleOptimizationAlgorithm,WOA)优化SVM区域高程异常拟合法。本方法利用WOA获取SVM中核函数与正则化参数,替代SVM算法中全局性搜索较差的交叉验证方法,构建高精度区域高程异常拟合模型。试验结果表明,采用WOA-SVM方法得到的高程异常拟合结果在精度、稳定性上均优于现有多项式拟合法和SVM拟合法,可为相关区域高程异常拟合提供一种有效方法。关键词:高程拟合;鲸鱼算法;支持向量机模型;核函数;正则化参数中图分类号:P237.2DOI:10.20007/j.cnki.61-1275/P.2023.02.13RegionalelevationanomalyfittingmethodbasedonSVMoptimizedbyWOAFANJianlin1,GAOYe1,YAOHui2(1.ZhejiangInstituteofSurveyingandMappingScienceandtechnology,Hangzhou,Zhejiang310000,China;2.NingBoMetallurgicalSurveyandDesignResearchCo.,Ltd.,Ningbo,Zhejiang315016,China)Abstract:InviewoftheproblemthatisdifficulttodeterminethebestparametersofregionalelevationanomalyfittingmethodbyusingSVM,thispaperproposesaregionalelevationanomalyfittingmethodbasedonSVMoptimizedbyWOA.WOAalgorithmisusedtoobtainkernelfunctionandregularizationparametersinSVM,replacethecrossvalidationmethodwithpoorglobalsearchinSVMalgorithm,andconstructahigh-precisionregionalelevationanomalyfittingmodel.ExperimentsshowthattheaccuracyandstabilityoftheelevationanomalyfittingresultsobtainedbyWOA-SVMmethodaresuperiortotheexistingpolynomialfittingmethodandSVMfittingmethod,fittingmodelarealsobetterthanthoseofpol-ynomialfittingmethodsandSVMfittingmethods,whichcanbeaneffectivemethodfortheheightanoma-lyfittingofrelevantregions.Keywords:elevationfitting;WOA;SVM;kernelfunction;regularizationparameter随着GNSS的快速发展,利用该技术获取地...