1、 l :2003-07-20; :2003-09-30 “:SE/ ?Z9(2002AA133060,2002AA130020,2002AA134090)bTe: Cf(1977- ), o, V 3,1V Y*b* sZE8 Z刘龙飞, 陈云浩, 李 京(北京师范大学资源科学研究所,北京师范大学资源信息科学与工程研究中心,北京 100875)K1:遥感影像的纹理分析已经成为一种重要的提高遥感影像分类精度的手段。着重介绍了用于遥感影像纹理分析的方法,对这些方法进行了分类和综合;这些方法的类别是:统计方法、结构方法、模型方法以及基于数学变换的方法。接着分别对各类别中的多种纹理分析方法进行了剖析,
2、列举各自的纹理特征,并指出了这些方法的优缺点和适应性。然后对应用这些方法的影像分类效果做了对比分析。最后分析了遥感影像纹理分析近年来的发展方向并对未来发展进行了展望。1oM:统计纹理分析;基于模型的纹理分析;数学变换纹理分析;纹理特征ms |: TP 75DS M :AcI|:1004-0323(2003)06-0441-071 引 言20 W90 M , “*/ 4,sO q*( IKONOS, SPOT5,COSMOS,OrbView) b“ V l bWS =4V%Mav1 *ma4 | aS Yb 7 bWsO q4i ?9 P.d; s 4q1rbs bWsO qs y 1,7 y
3、M : bWsO q PB 9F, h , By V Ps 4q2r;7 6BZ , bWsO q4 P1sO 9F,V7 P W; s9v,9v s 4, Ps by7 t f /, “ bWsO q4,s V ?q3,4rb B5B V %,1 =s0 ,lM “ us , t0 B ,ys TG bWsO q,9Gs q5r, s G 8s ZEb7G L; s , ZE94r s b j $ L80 TF,a aavl+,y7 V ATM0Fq6rbB s : ( V)d9 ( V)bLawsq7r/+ : (a aYa?5aLaZ_a qaMb Btid, q ,7Z_Z_b sa4
4、 | +, Vwr1; s Vsas C? 3b bWa; *, T bW 9s , * s V4q8rb sZE?s ZE,4N *Ks ,V7 Vz *,V* 4 |5b2 纹理分析方法 sZE VB ,:d9ZE,(+ )ZE, ZE# M(|) )ZEb/ * sZEbEo ps +, l , aZ_a, % VC Tb * 4 sOb7 Tf ) ,1 1 Cb )E b18 6 2003 M12*/ REMOTE SENSING TECHNOLOGY AND APPLICATIONVol.18No.6Dec. 20032. 1d9ZE2. 1.1 灰度共生矩阵法 V 3 g =,)
5、W bWs# bWM1“bHaralick1973 n54 3 (GLCM), Kn Cf:* sZE8 Z V1 _ ,YV, s b3 纹理分析方法比较 sZE,/ VBtZES B ,as1 bV3 ZE91;V4 Bt* sZEe1 b“ 4tZE,tZEV3 sZE1 ZE + 3 .da sZE, I n WM1“b 4, 3 3 ;7 , 3 3B“ +bE 1“d9b1M1 1o d( )3 g4bMarkov n51 Markov T, ,7 9 4, 6B Z(isotropic)bs Cssd, 4 usCbloME P (sO q)sZE, 6lof v4SbV4 sZ
6、EZE V UZE +BZmd9s4+ : (aSZa(ENT)Ms“ q16rbZmd9s +:= aQaM1aaQs axZq17rbs_ +: ? a (a= aQaaQs aZq18rbbW ro* HZE,YV ?_1f 0 0 _ ,YYVRobert 0aLaplacian 0; bW ro bW , Law ro,Law425 ro4 |, L, q7rb E +:(BWS)a+ (GS)aZ_a L a Cf:* sZE8 ZqJr. Pattern Recognition Letters, 1982,1(1): 43j50.q13rPaget R, Longsta I D,
7、 Lovell B. Texture ClassificationUs-ing Nonparametric Markov Random FieldqAr. 13th Interna-tional Conference on Digital Signal ProcessingqCr. 1998.q14rPentland A. Fractal-based Descriptionof Natural ScenesqJr.IEEE Transactions on Pattern Analysis and MachineIntelligence, 1984,9: 661j674.q15rChenC H,
8、 PauL F, Wang P S P. The Handbook of PatternRecognition andComputer Vision (2nd Edition)qMr. WorldScientific Publishing Co, 1998. 235.q16rGong P, Marceau D J, HowarthP J. A Comparison of Spat-ialFeature Extraction Algorithms for Land-use Classificationwith SPOT HRV DataqJr. Remote Sensing of Environ
9、ment,1992,40: 137j151.q17rHevenor R A. Third-order Co-occurrence Texture AnalysisApplied to Samples of High-resolution Synthetic ApertureRadar Imagery Engineering Topographic LaboratoryqRr. US Army Corps of Engineering, Fort Belvoir, VA, TechnicalReport, 1985. 32.q18rWeszka J S, Dyer C R, Rosenfeld
10、A. A Comparative Studyof Texture Measures for Terrain ClassificationqJr. IEEETransactions on Systems, Man and Cybernetics, 1976, 6:269j285.q19rLi W, He D C. A New Statistical Approach for TextureAnalysisqJr. Photogrammetric Engineering and RemoteSensing, 1990, 56: 61j66.q20rLi W. Automatic Identific
11、ation of Rocks on Thin SectionUsing Texture AnalysisqJr. Mathematical Geology, 1995,27: 847j865.q21rGanesan L, Bhattacharya P. A Statistical Design Of Exper-iments Approach For Texture DescriptionqJr. Pattern Reco-gnition Letters, 1995, 16: 99j105.q22rChenY Q, NixonM S, Thoma S D W. Statistical Geom
12、et-rical Features for Texture ClassificationqJr. Pattern Recog-nition, 1995,28(4): 537j552.q23rManian V, Vasquez R, Katiyar P. Texture ClassificationUsing Logical OperatorsqJr. IEEE Transaction on ImageAnalysis, 2000, 9(10): 1693j1703.q24rKovalev V, PetrouM. Multidimensional Co-Occurrence Ma-trices
13、for Object Recognition and MatchingqJr. GraphicalModels andImage Processing, 1996,58(3): 187j197.q25rValkealathi K, Oja E. Reduced Multidimensional Co-Occur-rence Histogramsin Texture ClassificationqJr. IEEETransa-ction on Pattern Analysis andMachine Intelligence, 1998, 20(1): 90j94.q26rI ,ym.t s ZE)qJr.Y,1997,(6): 38j41.q27r U , d y.s * ZE 3ms qJr.v(1 S),2000,36(6): 858j864.q28r 3,).s+ *m sZE#qJr. S/ f,1997,16(9 Cf:* sZE8 Z