GIS 390 GIS Vol.3, No., Summer 0 Iranian Remote Sensing & GIS 97-4 3 *...3 390// : 390// :.... Information ) SAM(Spectral Angle Measure) SCS(Spectral Correlation Similarity) Covariance-based Matched Filter ) JMD(Jeffiries-Matusita Distance) SID (Spectral Divergence RMFM(Correlation-based Matched Filter Measure) CEM (Constrained Energy minimizing) CMFM(Measure. System )) FIS (Fuzzy Inference System). (ANFIS (Adaptive Neuro-Fuzzy Inference CASI(Compact Airborne Spectrographic Imager) ANFIS. 0/9. ROC :.095569799 : : * Email: davoodakbari6@gmail.com
4 3..(Landgrebe, 999)... : 43. 000..(Chang, 000) AVIRIS 5 ( ) 00 6 7.(Chang,00) AVIRIS 8 (SAM).(Bakker, 00) HYMAP. Pattern Recognition. Image Space 3. Spectral Space 4. Feature Space 5. Mahalanobis Distance 6. Matched Filter Distance 7. Anomaly Detection 8. Spectral Distance Measures -..(Varshney, 004)...(Change, 003)....(Homayouni, 003) :. GIS 390 98
3. 5 4. 6. 7 Jeffries-. Matusita Mahalanobis..... Maximum likelihood classification. Receiver Operating Characteristic 3. Spectral Correlation 4. Deterministic Measure 5. Stochastic Measures 6. Spectral Matching 7. Spectral Similarity Value 003 (SCS) (CEM).(Homayouni, 003).(Lumme, 003) AISA 004 AVIRIS (SID) 006.(Du, 004) full-pixel sub-pixel. Cuprite Hymap NEVADA ROC. 007 AVIRIS.(Emami, 007).. 3.(Landgrebe, 999) GIS 390 99
.. -. 3 ROC. --.(Chang, 003).( ) ANFIS...... ANFIS.. ANFIS.. 3.. Boolian Operator. Euclidian Operator 3. Confusion Matrix GIS 390 00
s j s i L r j r i 4 3 TD ED CBD l l l.. (SAM) ( :(Yuhas, 99) (4) SAM cos (s i.s j/ s i s j ) L L / L / cos ( s l ils jl /[ s l il] [ s l jl] ) s j s i L r j r i. 90. 90 ( ). Spectral Signatures. City Block Distance 3. Euclidean Distance 4. Tchebyshev Distance.. ---. :..(Homayouni, 003) (SDM) ( l l l Chang, ) :(003 L CBD s l il s jl L l ED s, s [ ( s s i j TD max l L{ sil s jl } il jl ) ] () () / (3) GIS 390 0
() (). (SID) ( SID :(Chang, 003). (7) SID L L l p l log(p l / q l ) l q l log(q l / p l ) L L pl s il/ sik ql s jl/ s k k jk q l p l r j r i s jl. L s il. (JMD) Jeffries-Matusita ( Jeffries-Matusita :(Chang,003) (8) L JMD [ p l l q l] L L pl s il/ sik ql s jl/ s k k jk q l p l r i s jl s il. L r j. (SCS) (.(Homayouni,003) (5) L (s il s )(s l i jl s ) j SCS n s i sj s j s i L. n - +.. (SSV) ( l :(Homayuni, 003) SSV ED ( SCS) (6) SSV. [ 0 ] SSV.. SSV --- GIS 390 0
T (Ki K j) BD (si s j) (si s j) 8 (Ki K j) ln K i K j Ki Kj K (0) ( ) Mahalanobis Bhattacharyya. () T CMD (si s j) K L L(si s j) (Chang, 00) Chiang Chang () s i Mahalanobis ().. (CMD) KL L () ( 3 (Chang,00) RMD T RMD (si s j) R L L(si s j) s j RL L). () (MFD) ( s j s i.. Bhattacharyya Distance. Covariance-based Mahalanobis Distance 3. Correlation-based Mahalanobis Distance -3--.. (CEM) ( w d. L T yi wlril w ri l :(Harsanyi, 993) (9) w RL L d T d R L L d y r R L (Autocorrelation Matrix).. (MD)Mahalanobis ( Bhattacharyya ( K j K i ).(Landgrebe,999) s j s i GIS 390 03
(BO) ---. : (6) Boolianmap SAMmap SCSmap (ED) ---. ED (DN SAM ) (DN SCS ) : (7) ED. (FIS) -3-- IF- : OR AND.THEN NOT..... a b c d CMFM. (3) (Chang,00) T CMFM (s i ) K L L(s j ) (3) KL L. (3) RMFM ( RL L). 4 T RMFM si RL Ls j (Chang,00) (4) (MLC) (. () (5) T MLC ln K i (s i) K i (s i) K i i s. --... GIS 390 04
43-3-. commission. omission 3 (K) (OA) : 4 OA. OA. K.(Rosenfield, 986).. ROC.(Bradley, 997) 5 7 6. ROC. Neuro-Fuzzy. Overall Accuracy 3. Kappa coefficient 4. Mis-Matching 5. Detection power 6. Positive Probability 7. False Alarm Probability. (8). ANFIS -4-- ANFIS ANFIS. Jang(993)... GIS 390 05
8 0/4... 0 3 00. -4. 8 8..(-4 ).(3-4 ). ROC :.(3 ) ( ) ROC. -3... --3 CASI. CASI () ().3 GIS 390 06
() ( R=0.94, G=0.60, B=0.45) CASI ().4 (3) -. ED CBD TD SAM SCS SSV SID 0/886 0/880 0/900 0/96 0/96 0/938 0/963 0/60 0/546 0/65 0/847 0/85 0/779 0/85 0/4 0/0 0/00 0/039 0/038 0/06 0/037 0/04 0/06 0/033 0/05 0/06 0/08 0/08 JMD 0/96 0/853 0/038 0/0 CEM 0/966 0/873 0/034 0/ 05 CMD CMFM RMD RMFM MLC 0/99 0/968 0/904 0/966 0/99 0/694 0/880 0/665 0/873 0/694 0/08 0/03 0/096 0/034 0/08 0/04 0/08 0/033 0/05 0/04 --3 : () () ().ANFIS. 4 5. -.. GIS 390 07
(3) ( l ) () ( l ) ().5 (7) (6) (5) (4) ( l () ) Mahalanobis (0) (9) Jeffries-Matusita (8) (4) (3) Mahalanobis () (5).. (). 7. ANFIS.... - 6. GIS 390 08
(4) RMFM CMFM CEM (3) JMD SID () SCS SAM () (a).6 CMFM CEM (3) JMD SID () SCS SAM () (b). (5) JMD SID () SCS SAM () (c). (5) (4) RMFM SAM () ANFIS (d). (5) (4 ) RMFM CMFM CEM (3). (5) (4) RMFM CMFM CEM (3) JMD SID () SCS RMFM CMFM CEM JMD SID SCS SAM ANFIS ().7 () GIS 390 09
. - SAM,SCS 0/96 0/85 0/038 0/08 F.BO - SID, JMD 3- CEM,CMFM,RMFM 0/963 0/968 0/85 0/876 0/037 0/03 0/08 0/08 3 - SAM,SCS 0/96 0/96 0/84 0/850 0/039 0/038 0/035 0/06 F.ED - SID, JMD 3- CEM,CMFM,RMFM 0/963 0/966 0/853 0/874 0/037 0/034 0/08 0/04 3 - SAM,SCS 0/97 0/963 0/889 0/854 0/09 0/037 0/08 0/07 FIS - SID, JMD 3- CEM,CMFM,RMFM 0/96 0/968 0/853 0/879 0/038 0/03 0/0 0/05 3 - SAM,SCS 0/967 0/96 0/876 0/850 0/033 0/039 0/05 0/05 - SID, JMD 0/96 0/853 0/038 0/0 ANFIS 3- CEM,CMFM,RMFM 3 0/968 0/97 0/883 0/894 0/03 0/08 0/03 0/05 SAM,SCS, SID, JMD, CEM,CMFM,RMFM 0/973 0/900 0/07 0/0.... 9-3-3 SCS SAM JMD SID CMFM CEM RMFM 8.. 0/8 GIS 390 0
. ANFIS 0/9.. 0. CMFM. ANFIS. 0/883 0/880 0/894 ANFIS. ANFIS SAM SCS SID 0/900 JMD CEM CMFM RMFM Kappa Accuracy 0.905 0.900 0.895 0.890 0.885 0.880 0.875 0.870 CMFM-Anomaly Detection ANFIS CEM,CMFM,RMFM- Anomaly Detection ANFIS Two Steps. ANFIS- One Step.0 Kappa Accuracy 0.9 0.89 0.88 0.87 0.86 0.85 0.84 0.83 0.8 0.8 0.8 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0. 0. 0 ED CBD TD SAM SCS SSV SID JMD CEM CMD CMFM Spectral Detection Algorithm RMD.8 F.BO F.ED FIS ANFIS. RMFM - SAM,SCS - SID, JMD 3- CEM,CMFM,RMFM,,3.9.. ANFIS.. ANFIS MLC GIS 390
.. -5 Michel Roux. -6 Bakker, W.H. and Schmidt, K.S., 00, Hyperspectral Edge Filtering for Measuring Homogeneity of Surface Cover Types, ISPRS Journal of Photogrammetry & Remote Sensing 56, 46 56. Bradley, A.P., 997, The Use of the Area Under the ROC Curve in the Evaluation of Machine Learning Algorithms, Pattern Recognition, Vol.30, No.7, 45-59. Carvalho, O.A. and Meneses, P.R., 00, Spectral Correlation Mapper (SCM): An Improvement on the Spectral Angle Mapper (SAM), Asa Norte, 7090-900, Brasília, DF, Brasil. Chang, C.I, 003, Hyperspectral Imaging: Techniques for spectral Detection and Classification, Orlando, FL: Kluwer Academic. -4...... ANFIS. 90.. GIS 390
Chang, C.I. and Chiang, S.S., 00, Anomaly Detection and Classification for Hyperspectral Imagery, IEEE Trans. on Geoscience and Remote Sensing, vol.40, No.6. Chang, C.I, and Heinz D.C., 000, Constrained Subpixel Target Detection for Remotely Sensed Imagery, IEEE Trans on Geoscience and Remote Sensing, Vol. 38, NO. 3. Du, Y., Chang C.I. and Ren H., 004, New Hyperspectral Discrimination Measure for Spectral Characterization, Society of Photo-Optical Instrumentation Engineers. Emami, H., Afary, A., 007, Subpixel Classification on the Hyperspectral Images for Accuracy Improvement of Classification Results, Dep. of Geodesy and Geomatic Eng, K.N.Toosi University of Technology, Tehran-Iran. Farrand, W.H. and Harsanyi, J.C., 997, Mapping the Distribution of Mine Tailings in the Coeur d Alene River Valley, Idaho, Through the Use of a Constrained Energy Minimization Technique, Remote Sensing of Environment, No.59, 64-76. Frolov, D. and Smith, R.B., 999, Locally Adaptive Constrained Energy Minimization for AVIRIS Image, Eighth JPL Airborne Earth Science (AVIRS), http://www.microimages.com/papers/. Granahan, J.C. and Sweet, J.N., 00, An Evaluation of Atmospheric Correction Techniques Using The Spectral Similarity Scale, IEEE 00 International Geoscience and Remote Sensing Symposium, Vol. 5, 0-04. Harsanyi, J.C., 993, Detection and Classification of Subpixel Spectral Signatures in Hyperspectral Image Sequences, Ph.D. Dissertation, University of Maryland, Baltimore County, 6 pp. Homayouni, S. and Roux, M., 003, Material Mapping from Hyperspectral Images Using Spectral Matching in Urban Area, IEEE Workshop on Advances in Techniques for analysis of Remotely Sensed Data, NASA Goddard center, Washington DC, USA. Jang, J.S.R., 993, ANFIS: Adaptive-Networkbased Fuzzy Inference Systems, IEEE Transactions on Systems, Man, and Cybernetics, Vol. 3, No. 3, 665-685. Keranen, P., Kaarna, A. and Toivanen, P., 00, Spectral Similarity Measures For Classification in Lossy Compression of Hyperspectral Images, Proceedings of the 9'th International Symposium on Remote Sensing, SPIE, Vol.4885, Crete, Greece, September 00, 85-96. Landgrebe, D., 999, Some Fundamentals and Methods for Hyperspectral Image Data Analysis, SPIE Int. Symp. On Biomedical Optics (Photonics West), San Jose CA, Proc. SPIE Vol. 3603, 04-3. Lumme J.H., 003, Classification of Vegetation and Soil Using Imaging Spectrometer GIS 390 3
Data, Institute of Photogrammetry and Remote Sensing, Helsinki University of Technology, P.O.Box 00, FIN-005 HUT. Rosenfield, G.H. and Fitzpatric-Lins, K., 986, A Coefficient of Agreement as a Measure of Thematic Classification Accuracy, Photogrammetric Engineering and Remote Sensing, Vol. 5, No., 3-7. Varshney, P.K. and Arora, M.K., 004, Advanced Image Processing Techniques for Remotely Sensed Hyperspectral Data, Springer Berlin Heidelberg New York. Yuhas, R.H., Goetz, F.H. and Boardman, J.W., 99, Discrimination Among Semi-arid landscape Endmembers Using the Spectral Angel Mapper (SAM) Algorithm, In Summaries of the Third Annual JPL Airborne Geosciences Workshop, JPL publication 9-4. Vol. I. 47-49. CASI Technical Documentations, Itres Research Canada, http://www.itres.com/. GIS 390 4