基于PSO参数优化支持向量机的湿地遥感分类——以鄱阳湖部分区域为例
|
摘要
湿地遥感分类作为湿地管理和利用、动态监测的重要手段,是近年分类研究的热点之一,受到了广泛关注。鉴于湿地系统生态环境特殊,难以进行实地样本点的测取,选取小样本、多维度、高精度的分类方法显得尤为重要。以鄱阳湖国家自然保护区部分区域为研究范围,以landsat8 OLI影像为数据源,以支持向量机分类法为基础,通过PSO算法寻优SVM分类器的高斯核函数参数g和惩罚因子C,初步分析不同样本数目、光谱特征维度和辅助特征维度对于分类精度的影响,并同传统支持向量机分类方法进行比较。研究结果表明,在小样本、高维度情况下,基于PSO参数寻优的分类精度大于传统支持向量机分类精度。在样本数目达到100时,基于PSO参数寻优在高维度分类精度最佳,达到93.03%,较传统的SVM分类提高了1%左右。
Wetland classification of remote sensing dynamic monitoring of wetland management and utilization,important means,is one of the hotspots in the research of classification,has received the widespread attention.In view of the unique ecological environment of wetland system,field survey sampling this point,it is difficult to select small samples,the classification of multidimensional,high precision method is particularly important.Based on parts of the poyang lake national nature reserve as the research scope,to landsat8 OLI images as the data source,on the basis of support vector machine(SVM) classification,using PSO algorithm to optimize the SVM classifier of penalty factor C and the gaussian kernel function parameter g,wetland classification research,a preliminary analysisof the different sample size,spectral feature dimension and the auxiliary feature dimension as to the influence that the classification accuracy,and comparing with the traditional support vector machine(SVM) classification.The research results show that under the condition of small sample,high dimension,based on the classification of the PSO parameters optimization result accuracy than traditional support vector machine(SVM) classification accuracy.When the sample size of 100,based on the PSO parameters optimization in the highest classification accuracy,high dimension of 92.42%,up 3.22% compared with the traditional SVM classification.
Wetland classification of remote sensing dynamic monitoring of wetland management and utilization,important means,is one of the hotspots in the research of classification,has received the widespread attention.In view of the unique ecological environment of wetland system,field survey sampling this point,it is difficult to select small samples,the classification of multidimensional,high precision method is particularly important.Based on parts of the poyang lake national nature reserve as the research scope,to landsat8 OLI images as the data source,on the basis of support vector machine(SVM) classification,using PSO algorithm to optimize the SVM classifier of penalty factor C and the gaussian kernel function parameter g,wetland classification research,a preliminary analysisof the different sample size,spectral feature dimension and the auxiliary feature dimension as to the influence that the classification accuracy,and comparing with the traditional support vector machine(SVM) classification.The research results show that under the condition of small sample,high dimension,based on the classification of the PSO parameters optimization result accuracy than traditional support vector machine(SVM) classification accuracy.When the sample size of 100,based on the PSO parameters optimization in the highest classification accuracy,high dimension of 92.42%,up 3.22% compared with the traditional SVM classification.
引文
[1]刘蕾,臧淑英.兼容光学雷达影像及地形辅助数据的扎龙湿地遥感分类[J].地理与地理信息科学,2013,129(1):36-40.
[2]张策,臧淑英,金竺,等.基于支持向量机的扎龙湿地遥感分类研究[J].湿地科学,2011,99(3):263-269.
[3]董树斌,崔光范,朱宝光,等.洪河湿地三江平原生物基因库[M].佳木斯:黑龙江洪河国家级自然保护区,GEF湿地项目三江平原项目办公室,2007.
[4]Wright C,Gallant A.Improved wetland remote sensing in Yellowstone National Park using classification trees to combine TM imagery and ancillary environmental data[J].Remote Sensing of Environment,2007,107107(4):582-605.
[5]Zhang S Q,Na X,Kong B,et al.Identifying landscape pattern dynamics of Sanjiang plain marsh based on remote sensing techniques[J].Wetlands,2009,2929(1):302-313.
[6]Stehman S V,Wickham J D,Smith J H,et al.Thematic accuracy of the 1992 National Land-Cover Data for the eastern United States:Statistical methodology and regional results[J].Remote Sensing of Environment,2003,8686:500-516.
[7]Wickham J D,Stehman S V,Smith J H,et al.Thematic accuracy of the 1992 National Land-Cover Data for the western United States[J].Remote Sensing of Environment,2004,911:452-468.
[8]范昕炜.支持向量机算法的研究及其应用[D].杭州:浙江大学,2003.
[9]朱永生,张优云.支持向量机分类器中几个问题的研究[J].计算机工程与应用,2003,39(13):36-38.
[10]王振武.改进粒子群算法优化的支持向量机及其应用[J].哈尔滨工程大学学报,2016,7(12):1728-1733.
[11]Mitsch W J,Gosselink J G.Wetlands[M].2nded.New York:Van Nostrand Reinhold,1993:507-527.
[12]互动百科.http://www.baike.com/wiki/%E9%84%B1%E9%98%B3%E6%B9%96.
[13]Fisher B,Turner R K,Morling P.Defining and classifying ecosystem services for decision making[J].Ecological Economics,2009,6868(3):643-653.
[14]姚云军,张泽勋,秦其明,等.基于支持向量机的遥感影像湿地信息提取研究[J].计算机应用研究,2008,255(4):989-991.
[15]张树清,陈春,万恩璞.三江平原湿地遥感分类模式研究[J].遥感技术与应用,1999,1414(1):54-58.
[16]Li J,Chen W.A rule-based method for mapping Canada's wetlands using optical,radar and DEM data[J].International Journal of Remote Sensing,2005,26(22):5051-5069.
[17]Zhao H T,Jamie M.Data Mining with SQL Server 2005[M].New York:John Wiley&Sons,2006:6-10.
[18]Su L H.Optimizing support vector machine learning forsemiarid vegetation mapping by using clustering analysis[J].ISPRS Journal of Photogrammetry and Remote Sensing,2009,6464:407-413.
[19]Fogel D B.An introduction to simulated evolutionary optimization[J].IEEE Transactions on Neural Networks,1994,55(1):3-14.
[2]张策,臧淑英,金竺,等.基于支持向量机的扎龙湿地遥感分类研究[J].湿地科学,2011,99(3):263-269.
[3]董树斌,崔光范,朱宝光,等.洪河湿地三江平原生物基因库[M].佳木斯:黑龙江洪河国家级自然保护区,GEF湿地项目三江平原项目办公室,2007.
[4]Wright C,Gallant A.Improved wetland remote sensing in Yellowstone National Park using classification trees to combine TM imagery and ancillary environmental data[J].Remote Sensing of Environment,2007,107107(4):582-605.
[5]Zhang S Q,Na X,Kong B,et al.Identifying landscape pattern dynamics of Sanjiang plain marsh based on remote sensing techniques[J].Wetlands,2009,2929(1):302-313.
[6]Stehman S V,Wickham J D,Smith J H,et al.Thematic accuracy of the 1992 National Land-Cover Data for the eastern United States:Statistical methodology and regional results[J].Remote Sensing of Environment,2003,8686:500-516.
[7]Wickham J D,Stehman S V,Smith J H,et al.Thematic accuracy of the 1992 National Land-Cover Data for the western United States[J].Remote Sensing of Environment,2004,911:452-468.
[8]范昕炜.支持向量机算法的研究及其应用[D].杭州:浙江大学,2003.
[9]朱永生,张优云.支持向量机分类器中几个问题的研究[J].计算机工程与应用,2003,39(13):36-38.
[10]王振武.改进粒子群算法优化的支持向量机及其应用[J].哈尔滨工程大学学报,2016,7(12):1728-1733.
[11]Mitsch W J,Gosselink J G.Wetlands[M].2nded.New York:Van Nostrand Reinhold,1993:507-527.
[12]互动百科.http://www.baike.com/wiki/%E9%84%B1%E9%98%B3%E6%B9%96.
[13]Fisher B,Turner R K,Morling P.Defining and classifying ecosystem services for decision making[J].Ecological Economics,2009,6868(3):643-653.
[14]姚云军,张泽勋,秦其明,等.基于支持向量机的遥感影像湿地信息提取研究[J].计算机应用研究,2008,255(4):989-991.
[15]张树清,陈春,万恩璞.三江平原湿地遥感分类模式研究[J].遥感技术与应用,1999,1414(1):54-58.
[16]Li J,Chen W.A rule-based method for mapping Canada's wetlands using optical,radar and DEM data[J].International Journal of Remote Sensing,2005,26(22):5051-5069.
[17]Zhao H T,Jamie M.Data Mining with SQL Server 2005[M].New York:John Wiley&Sons,2006:6-10.
[18]Su L H.Optimizing support vector machine learning forsemiarid vegetation mapping by using clustering analysis[J].ISPRS Journal of Photogrammetry and Remote Sensing,2009,6464:407-413.
[19]Fogel D B.An introduction to simulated evolutionary optimization[J].IEEE Transactions on Neural Networks,1994,55(1):3-14.