遗传算法优化的支持向量机湿地遥感分类——以洪河国家级自然保护区为例

doi:10.13249/j.cnki.sgs.2012.04.434

摘要/Abstract

摘要：

湿地遥感分类作为湿地管理、监测与评价的重要手段,受到了广泛的关注。遗传算法(GA)借鉴了生物进化规律进行启发式搜索寻优,支持向量机(SVM)是一种新型的空间数据挖掘方法,二者相结合可以发挥各自的优势,寻找到支持向量机的全局最优参数,从而较准确地对湿地进行遥感分类。以洪河自然保护区为例,采用遗传算法优化的支持向量机方法进行了湿地遥感分类研究。同格网搜索下的支持向量机湿地遥感分类及最大似然监督分类对比,结果表明,遗传算法优化较格网搜索方式总精度提高了7.29%,较最大似然监督分类提高了12.06%,方法改善了沼泽、草地与裸地三种地物间的区分,是湿地遥感分类的有效手段。

关键词: 湿地, 遥感分类, 遗传算法, 支持向量机, 洪河自然保护区

Abstract:

Wetland remote sensing classification, as an important means of wetland management, monitoring and assessment, has been widely concerned. Genetic Algorithm (GA) does heuristic search optimization which references the law of biological evolution, while Support Vector Machine (SVM) is a new kind of spatial data mining method. Combination of both can develop their own advantages to do wetland remote sensing classification exactly, by searching the global optimal parameters of Support Vector Machine. Taking Honghe Nature Reserve as a case study, wetland remote sensing classification using Support Vector Machine optimized with Genetic Algorithm (GA-SVM) was explored in this paper. In comparison with wetland classification using support vector machine with parameters searched by Grid and the Maximum Likelihood Classification. The experimental results show that, the overall accuracy of Genetic Algorithm optimization has increased 7.29% compared to Grid Search method, and has increased 12.06% compared to the Maximum Likelihood Classification, by improving the discrimination among marsh, meadow and bare land. Therefore, GA-SVM is an effective tool in wetland remote sensing classification.

Key words: wetland, remote sensing classification, Genetic Algorithm, Support Vector Machine, Honghe National Nature Reserve

中图分类号:

TP751.1

臧淑英, 张策, 张丽娟, 张玉红. 遗传算法优化的支持向量机湿地遥感分类——以洪河国家级自然保护区为例[J]. 地理科学, 2012, 32(4): 434-441.

Shu-ying ZANG, Ce ZHANG, Li-juan ZHANG, Yu-hong ZHANG. Wetland Remote Sensing Classification Using Support Vector Machine Optimized With Genetic Algorithm: A Case Study in Honghe Nature National Reserve[J]. SCIENTIA GEOGRAPHICA SINICA, 2012, 32(4): 434-441.

图/表 7

图1

表1

表2

表3

表4

图2

图3

参考文献 27

[1]	Mitsch W J, Gosselink J G.Wetlands[M]. 2nd ed. New York: Van Nostrand Reinhold, 1993: 507-527.
[2]	Fisher B, Turner R K, Morling P.Defining and classifying ecosystem services for decision making[J]. Ecological Economics, 2009, 68(3): 643-653.
[3]	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): 5 051-5 069.
[4]	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, 29(1): 302-313.
[5]	Özesmi S L, Bauer M E.Satellite remote sensing of wetlands[J]. Wetlands Ecology and Management, 2002, 10(5): 381-402.
[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, 86: 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, 91: 452-468.
[8]	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, 107(4): 582-605.
[9]	Zhao H T, Jamie M.Data Mining with SQL Server 2005[M]. New York, USA, John Wiley & Sons, 2006:6-10.
[10]	Su L H.Optimizing support vector machine learning for semi-arid vegetation mapping by using clustering analysis[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2009, 64: 407-413.
[11]	Fogel D B.An introduction to simulated evolutionary optimization[J]. IEEE Transactions on Neural Networks, 1994, 5(1): 3-14.
[12]	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, 25(4): 989~991.
[13]	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, 9(3): 263~269.
[14]	Li S J, Wu H, et al.An effective feature selection method for hyperspectral image classification based on genetic algorithm and support vector machine[J]. Knowledge-Based Systems, 2011, 24(1): 40-48.
[15]	赵东升,吴正方,商丽娜.洪河保护区湿地生态需水量研究[J].湿地科学,2(2):133~138.
[16]	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.
[17]	Vapnik V.The Nature of Statistical Learning Theory[M]. New York: Springer-Verlag, 1995.
[18]	Melgani F, Bruzzone L.Classification of hyperspectral remote sensing images with support vector machines[J]. IEEE Trans Geosci Remote Sens., 2004, 42(8): 1 778-1 790.
[19]	Foody G M, Mathur A.A relative evaluation of multiclass image classification by support vector machines[J]. IEEE Trans Geosci Remote Sens, 2004, 6(42): 1335-1343.
[20]	Pal M, Mather P M.Support vector machines for classification in remote sensing[J]. Int J Remote Sens, 2005, 3(26): 1007-1011.
[21]	Chang C C, Lin C J. LIBSVM: A library for support vector machines [EB/OL]. Software available at , 2001.
[22]	Holland J H.Adaptation in Natural and Artificial Systems [M]. Ann Arbor: The University of Michigan Press, 1975.
[23]	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, 14(1): 54~58.
[24]	Haralick R M, Shanmugam K, Dinstein I.Texture features for image classification[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1973, 3: 610-621.
[25]	Zar J H. Biostatistical Analysis (2nd Edition)[M]. Englewood Cliffs: Prentice-Hall, 1984.
[26]	骆剑承, 王钦, 马江洪,等. 遥感图像最大似然分类方法的EM改进算法[J]. 测绘学报, 2002, 31(3): 234~239.
[27]	Thomas O, Debasmita M, Navin K C T, et al. An objective analysis of support vector machine based classification for remote sensing[J]. Math Geosci., 2008, 40: 409-424.

湿地	水田	指有水源保证和灌溉设施,在一般年景能正常灌溉,用以种植水稻,莲藕等水生农作物的耕地,包括实行水稻和旱地作物轮种的耕地。
湿地	沼泽	指地势平坦低洼,排水不畅,长期潮湿,季节性积水或常积水,表层生长湿生植物的土地。
非湿地	耕地	指无灌溉水源及设施,靠天然降水生长作物的耕地;有水源和浇灌设施,在一般年景下能正常灌溉的旱作物耕地;以种菜为主的耕地,正常轮作的休闲地和轮歇地。
	裸地	目前还未利用的土地、包括难利用的土地。
	村镇	指城乡居民点及县镇以外的工矿、交通等用地。
	草地	指以生长草本植物为主,覆盖度在5%以上的各类草地,包括以牧为主的灌丛草地和郁闭度在10%以下的疏林草地。
	林地	指生长乔木、灌木、竹类、以及沿海红树林地等林业用地。

样点数	水田	耕地	沼泽	裸地	村镇	草地	林地	总数	用户精度
水田	235	4	0	0	0	29	0	268	87.69%
耕地	3	276	0	0	0	0	0	279	98.92%
沼泽	0	0	234	0	17	0	0	251	93.23%
裸地	0	2	19	224	0	47	5	297	75.42%
村镇	0	0	8	2	215	0	0	225	95.56%
草地	0	4	7	21	0	229	38	299	76.59%
林地	0	0	3	0	0	32	253	288	87.85%
总数	238	286	271	247	232	337	296	1907
制图精度	98.74%	96.50%	86.35%	90.69%	92.67%	67.95%	85.47%
总体精度 =87.36% Kappa系数 =0.852

样点数	水田	耕地	沼泽	裸地	村镇	草地	林地	总数	用户精度
水田	224	42	0	0	0	5	0	271	82.66%
耕地	14	236	0	0	0	20	0	270	87.41%
沼泽	0	0	210	0	14	0	0	224	93.75%
裸地	0	4	3	200	0	99	4	310	64.52%
村镇	0	0	22	1	217	0	0	240	90.42%
草地	0	4	29	46	1	174	26	280	62.14%
林地	0	0	7	0	0	39	266	312	85.26%
总数	238	286	271	247	232	337	296	1907
制图精度	94.12%	82.52%	77.49%	80.97%	93.53%	51.63%	85.26%
总体精度 =80.07% Kappa系数 =0.767

样点数	水田	耕地	沼泽	裸地	村镇	草地	林地	总数	用户精度
水田	221	41	0	0	1	26	0	289	76.47%
耕地	0	230	2	10	0	0	0	242	95.04%
沼泽	0	2	198	6	22	11	9	248	79.84%
裸地	15	3	46	165	0	59	32	320	51.56%
村镇	0	1	0	0	203	0	0	204	99.51%
草地	2	6	14	66	6	187	23	304	61.51%
林地	0	3	11	0	0	54	232	300	77.33%
总数	238	286	271	247	232	337	296	1907
制图精度	92.86%	80.42%	73.06%	66.80%	87.50%	55.49%	78.38%
总体精度 =75.30% Kappa系数 =0.712