摘要
森林是国家可持续发展的重要物质基础，是经济建设和生态环境建设中不可多得的、可更新的再生资源，具有保持水土、涵养水源、防风固沙、净化空气、调节气候等生态功能，森林资源状况及其消长变化，不仅影响区域经济的持续发展，而且还影响地区乃至全球环境的变化，因此倍受人们关注。
遥感技术具有宏观、动态、便捷、可周期重复和成本低等诸多优点，非常吻合森林资源辽阔性、复杂性、通达性差等特点，已成为研究森林资源状况的理想手段。目前，在森林遥感应用中，多源遥感数据的提供能力越来越强，但由于遥感信息的综合性、复杂性，使得遥感信息处理技术却相对落后。基于不同区域、不同季相和不同背景特征的森林遥感分类技术远未成熟，高分辨率IKONOS卫星影像在林业上的应用研究更是少见报道。
本文以浙西北山区(淳安和临安两个研究点)为研究对象，利用IKONOS和TM遥感数据针对森林遥感分类中的影像处理、波段组合选择、多源影像融合、影像分类方法和森林遥感分类效益等问题开展研究，对推进浙江地区森林遥感应用具有重要意义。
研究的主要内容和结论归纳如下：
(1)分别对研究区TM和IKONOS影像进行了各波段信息量、标准差、相关性等统计特征及森林类型光谱特征分析。通过定性分析和最佳指数因子OIF值计算，表明，TM541、IKONOS421是一种最佳的三波段组合方式，具有最大的信息量和最少的信息冗余度。研究还对两种影像做了主成分分析，提取出主成分特征影像。
(2)研究进行了HIS、Brovey、主成分变换等多种影像融合试验，通过融合前后影像的目视比较和相关系数定量分析得出，主成分变换法是一种优良的影像融合方法。融合后的IKONOS影像，既具有丰富的色彩，又具有清晰的纹理结构，有利于森林类型分类。
（3）在 GIS数据支持下，进行了多种森林遥感分类试验，精度分析表明：
对于TM影像来说，基于原始六波段影像（热波段除外）的监督分类效果最佳，
但若分至四级森林类型，其分类总体精度只能达 70．67％，面积相对精度仅为
60．13％，与实用要求相去甚远，如果分至二级，则面积精度可以达到 88．63％。
在监督分类中，研制了一种专题矢量图件与种子像元扩展紧密结合的训练样本优
选法，可使训练样本得以优化和纯化。对于 IKONOS高分辨率融合影像，人机交
互目视解译是一个最佳方法，森林四级类型的目视解译面积精度可达 97．60％。
门）研究表明 IKONOS适宜用于森林面积资源本底清查，完全满足制作
．：5000 山林现状图的精度要求，并具有良好的经济效益，每平方公里林业用地
可节省经费约70元，整个浙江省完成一次森林资源调查，可节省经费约458万
元（不考虑遥感数据的其它综合利用价值）。
Forest resource is the important material foundation of national sustainable development. It is the rare , renewable resources in the economic construction and ecological environment construction, having the ecological function of water and soil conserving,wind-defensing,sand-binding, air purifying and weather egulating, etc. The forest resource state of growth and decline influence not only the sustainable development of the regional economy , but also the changes of region and even global environment.Therefore,it has attracted attention extremely.
Possessing the advantages of macroscopic information, realtime,convenience,periodicity and low cost, remote sensing is suitable for forestry with the features of extensity,complexity and difficulty to access, and has already become the ideal means to study forest reserves state. At present, in the application of forest, the ability to gain multi-sources remote sensing data is stronger and stronger, but because of the integrity of remotely sensed information and the mechanism complexity of remote sensing, information process technology lag behind information acquisition technology relativly. It is far away ripe for the forest types classification by remote sensing basis of different areas , seasonal aspect and different background characteristics. The application study on the forestry of the high-resolution IKONOS satellite image has been reported rarely.
So, this research regards mountain area of the northwest of Zhejiang( two cases in Chunan and Linan are clicked) as the research objects, some key problems are discussed about image processing, band combination selection, image merge, image interpretation, and the efficiency of forest classification by remote sensing. It will be significant effect to promote the forest remote sensing application in Zhejiang.
The main content and conclusion of this research is summed up as follows:
(1) Analyzing TM and IKONOS images amount,standard deviation and the correction between bands. Meantime, analyzing spectrum features of main forest types. Based on various qualitative analysis and quantitative calculation of the value of OIF , the different band combination means have been discussed and TM541 and IKONOS421 have been proposed in forest classification because of possessing the largest amount of information and the least redundancy .The feature images were extracted through principal component transformation.
(2) Different merge methods such as HIS, Brovey, and principal component transformation have been discussed. Based on correlation coefficient quantitative analysising, conclution was given that principal component transformation is a kind of fine merge method .The merged IKONOS image posses both spectrum information of the multi-spectrum image and structure information of the high resolution image, which is benefit to forest types classification.
(3) With the support of GIS data ,various forest classification technology by remote sensing have been tested . The results indicate that supervised classification based on original six bands images ( except the hot wave band) have the highest accuracy for TM image .But if forest types were classified to the fourth level, The overall accuracy can only be up to 70.67% and the area relative precision is only 60.13%. If were classified to the second level, the area relative precision may up to 88.63% then. In supervised classification, an optimal selection method of training sample that combine thematic vector and seed pixel expanding tightly has been proposed , which can optimize and purify the training sample. As to the IKONOS image ,the human- computer interaction interpretation is most effective, the area relative precision of which can be up to 97.60%.
(4) Research indicate IKONOS image is suitable for using in forest resource inventory , which meets the precision request totally of producing 1:5000 forest map and will be good economic benefits. It can save about 70 yuan every sq. km forest land and about 4,580,000 yuan in the whole Zhejia
引文
[1]徐冠华，田国良等．遥感信息科学的进展与展望[J]．地理学报，1996，51(5)：385～397．
[2]梅安新，彭望禄，秦其明等．遥感导论[M]．北京：高等教育出版社，2001．
[3]林辉，童显德，黄忠义．遥感技术在我国林业中的应用与展望[J]．遥感信息，2002(1)：39～43．
[4]郭华东．对地观测技术与可持续发展[M]．北京：科学出版社，2001．
[5]Dao Huiwu. Amazone River Basin-the Bad Lung of the Earth [N]. The World News,2001,1,14(15).
[6]联合国气候变化框架公约京都议定书[EB/OL].http://www.Unfccc.de/resource/docs/convkp/kpchinese.Pdf.1997.
[7]United Nation Framework Convention on Climate Change[J/OL]. retrieved from: http://www.Unfccc.de/resource.conv/index.Htm.1997.
[8]Karl M,R Paivinen. Definition of a System of Nomenclature For Mapping European Forests and for Compiling a Pan—European Forest Information System[C]. Office for official publications of the European Communities, Luxembourg, 1996.
[9]Lin C, R. Pivinen. Forestry user requirement assessment the SAR for agriculture and forestry in Europe(SAFE)project experiences and results[C]. In: Proceedings of the 2nd international workshop on Retrieval of Bio—and Geo—physical parameters from SAR data for land applications, 21—23 October, 1998,Eds: Borgeaud M. and T. Guyenne, ESTEC, the Netherlands, 1998. SP—411, 614pp.
[10]李秀彬．全球环境变化研究核心领域—土地利用／土地覆盖变化的国际研究动向[J]．地理学报，1996，51(67)：553—557．
[11]郑小贤．德国、奥地利和法国的多目的森林资源监测述评[J]．北京林业大学学报，1997，19(3)：79—85．
[12]Maribeth H Price. Application of Remote Sensing to Forest Resource Inventory and Habitat Modeling[J]. NASA NRA—98—OES—09, 2000.
[13]01ga Rigina, H. kan 01sson. Detection of Boreal Forest Decline by High—Res01ution Satellite lmagery, Conference Proceedings of Remote Sensing and Forest Monitoring[J/OL]. retrieved from: http://rogow99.sggw.waw.pl/.1999.
[14]CP Gross. Contributions of Remote Sensing to Forest Health Monitoring in Europe beyond 2001. Conference Proceedings of Remote Sensing and Forest Monitoring [J/OL]. retrieved from: http://rogow99.sggw.waw.pl/.1999.
[15]Kuntz S, Siegert F. Monitoring of deforestation and land use in Indonesia with multitemporal ERS data[J]. International journal of Remote Sensing, 1999, (20): 2835—2853.
[16]Haripriya, G.S. Estimates of biomass in Indian forests[J].Biomass and Bioemergy,2000,19(4): 245-258.
[17]Kaygusuz K, Biomass energy potential in Turkey. Renewable Energy[J] .2002,26(4): 661-678.
[18]Schroeder, P. Biomass estimation for temperate broadleaf forests of the United States using
inventory data[J] .Forest Science,1997,42(3): 424-434.
[19]Ravindranath, N.H.(1997), Carbon flows in Indian forest. Climatic Change. pp. 297-320.
[20]Ramsey Ⅲ, Elijah W.; Hodgson, Michael E.; Sapkota, Sijan. K. Forest impact estimated with NOAA AVHRR and Landsat TM data related to an empirical hurricane wind-field distribution[J]. Remote Sensing of Environment ,2001,77(3): 279-292.
[21]史培军，宫鹏，李晓兵等，土地利用／覆盖变化研究的方法与实践[M]．科学出版社，2000，47-51．
[22]武红敢．“3S”技术在美国林业研究中的最新进展及其应用[J]．世界林业研究，1998，3：52-57．
[23]McDonald, A. J., Gemmell, F. M. and Lewis, P. E., Investigation of the utility of spectral vegetation indices for determining information on coniferous forests[J]. Remote Sensing of Environment. 1998,66(2): 250-272.
[24]Verstraete, M. and Pinty, B. (1996), Designing optimal spectral indexes for remote sensing applications[J]. IEEE Trans. Geosci. Remote Sens. pp. 1254-1265.
[25]D. Wu and J. Linders(2000), Comparison of three methods to select feature for discriminating forest cover types using SAR imagery[J]. INT. J. Remote Sens. 2000,21(10):2089-2099.
[26]Mayaux Philippe, De Grandi Gianfranco. Central African Forest Cover Revisited[J]. Remote Sensing of Environment, 2000, 71(2): 183-196
[27]Martin, M. E.,Newman, S. D. and Aber, J. D. Determining Forest Species Composition Using High Spectral Resolution Remote Sensing Data[J].Remote Sensing of Environment, 1998,65(3): 249-254.
[28]Atkinson, P.M. and Tatnal, A.R.L., Neural networks in remote sensing, Int. J. Remote sensing, 1997,18(4):699-709
[29]谭宽祥．卫星遥感与森林资源监测[J]．中南林业规划，1995，51：(1)48-51．
[30]林辉，童显德，黄忠义．遥感技术在我国林业中的应用与展望[J]．遥感信息，2002，1：39-43．
[31]章杨清，刘政凯．利用分维向量改进神经网络在遥感模式识别中的分类精度[J]．环境遥感，1994，9(1)：68—72．
[32]靳文戟，刘政凯．多类别遥感图象的复合分类方法．环境遥感，1995，10(4)：288-301．
[33]吴炳方，黄绚，田志钢．应用遥感及地理信息系统进行植被制图．环境遥感，1995，10(1)：30—36．
[34]李崇贵等．以遥感和地理信息系统为基础的森林蓄积LS估计自变量的选择研究[J]．遥感学报，2001，5(4)：277—281．
[35]邓国芳，蒙彬．广西河池地区利用航天遥感技术进行森林资源调查技术研究[J]．中南林业调查规划，2001，20(1)：14-17．
[36]杨成华，安和平．贵州南北盘江流域植被类型的卫片解译与制图[J]．贵州林业科技，1996，24(1)：55-58．
[37]杨彪．TM影像目视判读在“两项调查”中的应用[J]．云南林业调查规划设计，2000，
25(3): 22-25.
[38]Pohl c., Van Genderen J. L. Mulfisensor image fusion in remote sensing: concept, methods and applications. INT. Remote sensing, 1998, 19(5): 823-854.
[39]Van Genderen J.L.,Pohl C. Image fusion: Issue, techniques and applications. Preceedings EARSeL Workshop on Intelligent Image Fusion, Van Genderen J.L. And Cappellini V. (EDS), Holiday Inn, Strasbourg, France, 11 September 1994.
[40]Chavez P. S. Jr., Sides S.C., Andenson J.A. Comparison of three different methods to merge multiresolusion and multispectral data: TM & SPOT pan. Photogrammetric Engineering and Remote Sensing,1991,57(3):295—303.
[41]Leckie D. G Syergism of SAR and visible/infrared data for forest type discrimination[J]. Photogrammetric Engineering and Remote Sensing, 1990,56(9): 1237—1246.
[42]Perlant F, Sempere J. P. and Guerre L. F. Production of SPOT/ERS-1 image maps. Proceedings First ERS-1 Pilot Project Workshop, Toledo, Spain,22—24 June, ESA SP-365,1994,331—336.
[43]Strobl D, Raggam J, Buchroithner M. F. Terrain correction geocoding of a multisensor image data set[J]. Proceeding 10th EARSeL Symposium,Toulouse,France,1991,98—107.
[44]Sheffigara V. K. A generalized component substition technique for spatial enhancement of multispectral images using a higher resolution data set[J]. Photogrammetric Engineering and Remote Sensing, 1992,58(5): 561-567.
[45]孙家柄，刘继林，李军．多源遥感影像融合[J]．遥感学报，1998，2(1)：47-50．
[46]贾永红，李德仁，孙家柄．多源遥感数据融合[J]．遥感技术与应用，2000，15(1)：41-44．
[47]贾永红，李德仁，孙家柄等．四种IHS变换用于SAR与TM影像复合的比较[J]．遥感学报，1998，2(2)：103—106．
[48]李军，周月琴，李德仁．小波变换用于高分辨率全色影像与多光谱影像的融合研究[J]．遥感学报，1999，3(2)：116-121．
[49]孙永军，王志林．IKONOS卫星数据在天津市土地利用动态遥感监测中试验研究．第一届IKONOS北京技术研讨会，2001，6：20—25．
[50]徐希孺，周莲芳．混合像元的因子分析方法及其在大范围冬小麦播种面积估算中的应用探讨．科学通报，1989，34(12)：946-949
[51]阎静，王汶，季湘阁．利用神经网络方法提取水稻种植面积—以湖北省双季早稻为例．遥感学报，2001，5(3)：227-230
[52]赖格英，姚克敏．遥感影像混合象元的分析和处理．气象教育与科技，1998，20(2)：18-23
[53]王喜鹏，张养贞．应用神经网络模型分解AVHRR混合像元．遥感学报，1998，2(1)：51-56
[54]梅安新，彭望禄，秦其明等．遥感导论[M]．北京：高等教育出版社，2001．213-223
[55]王人潮，史舟，胡月明．浙江红壤资源信息系统的研制与应用[M]．北京：中国农业出版社，1999
[56]刘纪远，庄人方，凌杨荣．基于GIS的中国东北植被综合分类研究[J]．遥感学报，1998，1(3)：285-291
[57]刘纪远．国家资源环境遥感宏观调查与动态监测研究[J]．遥感学报，1997，1(3)：225—230
[58]浙江省林业厅．浙江省森林资源规划设计调查技术操作细则[M]．1997
[59]Erdas, INC. Erdas Field Guide. Fourth Edition. Atlanta, Georgia, USA, 1997
[60]Erdas, INC. Erdas Imagine Tour Book. Atlanta, Georgia, USA, 1999
[61]刘勇，贺雪鸿译．遥感精解．测绘出版社，1993
[62]许殿元，丁树柏．遥感图像信息处理．宁航出版社，1990
[63]张永生．遥感图像信息系统．北京，科学出版社，2000，4
[64]Chavez. P.S., Jr. An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data. Remote Sensing of the Environment. 1998, 24:459-479
[65]Chavez, P.S., Jr. G. L. Berlin, and L.b. Sowers. Statistical method for selecting Langsat MSS ratios. Journal of Applied Phototgraphic Engineering 1982, 8(1): 23-30
[66]陆灯盛，游先祥，崔赛华．TM图像的信息分析及特征信息提取的研究．环境遥感．1991，6(4)：282—292
[67]Congalton R.G, A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment. 1991, 37:35-46
[68]Congalton R.G, Mary Balogh, et al, Mapping and monitoring agricultural crops and other land cover in the lower Colorado River Basin, Photogramm. Eng. Remote Sens., 1998, 64(11): 1107-1113
[69]Janssen L.L.F. and Van Der Wel F. J. M., Acuracassessment of satellite derived land-cover data: A eview. PE&RS, 1994, 66(4): 419-426
[70]ChavezP.s.Jr., Sides S.C., Anderson J. A. Comparison of three different methods to merge multiresolution and multispectral data: TM & SPOT pan. Photogrammetric Engineering and Remote Sensing, 1991, 57(3):295-303.
[71]Sheffigara V.K.A generalized component substitution technique for spa-tial enhancement of multispectral images using a higher resolution data set. Photogrammetric Engineering and Remote Sensing, 1992, 58(5):561-567.
[72]Ydson H., Besnus Y., Polet J. Extraction of spectral information from Landsat TM data and merger with SPOT panchromatic imagery—a con-tribution to be study of geological
structures. ISPRS Journal of Pho-togrammetry and Remote Sensing, 1993,48(5):23-36.
[73]Yocky, D.A., Lmage merging and data fusion using the discrete two-di-mensional wavelet transform. Joumal of Opt. Soc. Am. A., 1995,12(9):1834-1841.
[74]A. Goshtasby, G.C.Stockman, C. V. Page. A region-based approach to digital image registration with subpixel accuracy. IEEE Trans. On Geoscience and Remote, 1986, GE-24(3).
[75]浙江省林业局．浙江林业自然资源森林卷[M]．北京：中国农业科学技术出版社，2002