SlideShare a Scribd company logo
IGARSS 2011, Vancouver
TU4.T08.1: Hyperspectral Monitoring of the Environment I




A Methodology of Forest Monitoring from
Hyperspectral Images with Sparse Regularization

                                                           Jul. 26, 2011


MITSUBISHI RESEARCH INSTITUTE, INC.

Keigo YOSHIDA, Takashi OHKI, Masahiro TERABE, Hozuma SEKINE (MRI)
Tomomi TAKEDA (ERSDAC)



Copyright (C) Mitsubishi Research Institute, Inc.
Introduction:Forest Monitoring by Remote Sensing

                                                       Decision making for Forest Management
                                                       Disaster prevention planning
                          Accurate Info.
    Needs                                              Finding forest of poor growth
    Needs                of Present Forest
                                                       GHG Credit estimation
                                                       Resource management

                                                     Forest conditions change dynamically
                            High Cost of
                                                     Conduct periodical field survey
  Problem                   Forest Survey
  Problem                                            estimate 600-1200 USD / Km2 / year
                            & Monitoring
                                                        ※ in case of Japan; 1 USD = 81 yen




           Forest Monitoring                         No need for field survey all over the area
  Solution
  Solution by Remote Sensing
                                                     Highly-frequent observation



Copyright (C) Mitsubishi Research Institute, Inc.         2
Introduction:Highly-developed Sensing Tech. & Challenges

              Sensing Tech.                                            Challenges
                                                     Hard to bring out potential of big sensor data
 Hyperspectral sensor                                 [e.g.] NDVI use just 2 bands, or Red and IR
                                                            & have to select optimal band combinations
 provide detailed optical info.
  on forest physiognomy           Complexity of prediction model increases,
    growth situation              resulting in poor prediction performance
      character of tree species                       Dimension is high but sample size is small
       etc.                                             due to limitation of field survey
                                                       This causes model overfitting

 Sensor fusion                                       Modeling is not easy for several sensor data
 Reflect diverse property                            of different physical property
  of targets                                         Statistical or Data-driven approach is needed


Copyright (C) Mitsubishi Research Institute, Inc.          3
Research Outline

 Utilize rich data by a machine learning technique (sparse regularization)
     and achieve accurate, informative, & less costly forest monitoring

                                      Remote Sensing Data Fusion
          Input Data                  (CASI-3 hyperspectral images + SAR signals)
          Input Data
                                      Field Survey Results

                                      Sparse Regularization
        Methodology
        Methodology                  (Sparse Discriminant Analysis、LASSO regression)

                                      Predicted Stand Factors of each subcompartments
                                       for Forest Management
         Output Data
         Output Data                 (Species, Canopy cover, Timber volume, Tree height)
                                      Prediction Models


    Subcompartment: a general spatial unit for forest monitoring

Copyright (C) Mitsubishi Research Institute, Inc.          4
Target Site


      Town-owned forest in Shimokawa,
             Hokkaido, Japan
 Approx. 90 % of town is covered by forest
 Utilize local conifer resources for business
 Environmental model city for low-carbon society




                                                Shimokawa




Copyright (C) Mitsubishi Research Institute, Inc.           5
Remote Sensing Data
      Species and Canopy-cover prediction:                    CASI-3 hyperspectrum
      Volume and Height prediction:                           Data fusion (CASI-3 + PALSAR)

 Remote Sensing
    Hyperspectral sensor                           (optical property)
                Airborne hyperspectral imager CASI-3
                84 bands from 400 to 1060 nm (wavelenght res. : 8 nm)
                Original spatial res.: 2.0 m
                 → Resolution is decreased to 30m to simulate satellite-based operation

        PALSAR         (shape or volume property)
                Microwave backscattering         org. image                        resized




Copyright (C) Mitsubishi Research Institute, Inc.         6
Field Survey

       During aircraft obs., conduct field survey to collect data for modeling & validation


 Field Survey:
           Place 25-sq-m quadrats
           Inventory study for trees whose DBH > 5cm & and record tree species
           Canopy cover measurement with whole-sky camera
           Height measurement for sampled 10 trees




Copyright (C) Mitsubishi Research Institute, Inc.   7
What is Sparse Regularization ? Why Do I Use it ?

   “Sparse” means the model has a low # of nonzero parameters

    ■ Optimal Band Selection

                    Ineffective parameters will be removed from prediction model
                      automatically by solving convex optimization problem


    ■ Higher Generalization Capability

                    simple model with smaller # of bands achieves less
                      overfitting; better prediction performance

    ■ More Interpretable Model

Copyright (C) Mitsubishi Research Institute, Inc.   8
Sparse Regularization:Theoretical Overview
      Add penalty to loss function to obtain model with small num. of variables

 LASSO (R. Tibshirani et al., 96)                          Loss function (LS)



                                                    (penalty)       norm

 Optimal Scoring (T. Hastie et al., 94)
       Perform Fisher’s linear discriminant analysis as regression by score
                convert categorical variables for class membership into quantitative
       Optimize                    and weight vector               simultaneously




Copyright (C) Mitsubishi Research Institute, Inc.               9
Intuitive Explanation of Sparse Regularization
 To reduce empirical errors,                       <penalty>
  W moves away from 0,
  then penalty increases


 L1-norm: attraction force to 0 is const.
  -> Small values in W tend to be 0
                                                    <attraction force to 0>
   L2-norm: attraction force is small around 0
   -> Small values in W remain


 Coefficients
                          L1-regularization                            L2-regularization




Copyright (C) Mitsubishi Research Institute, Inc.   10
Experimental Flow
1. Modeling                                                                Prediction
                                                                            Prediction
                                                                          Performance
                                                                          Performance
     Hyperspectral Reflectance
     Hyperspectral Reflectance                                            Classification
                                                                          Classification
                                                        Sparse LDA
      (ave. w/in each quadrat)
      (ave. w/in each quadrat)                                                Model
                                                                              Model
                                                                           Regression
                                                                           Regression
               PALSAR Signals
               PALSAR Signals                        LASSO Regression
                                                                              Model
                                                                              Model

 2. Prediction for Subcompartments
              Forest Pixel Extraction

      Hyperspectral Reflectance
      Hyperspectral Reflectance
                                                    Semisupervised LDA    Forest Pixels
                                                                          Forest Pixels
        (30m x 30m pixels)
         (30m x 30m pixels)

         Subcompartment Prediction



           Averaged Reflectance
           Averaged Reflectance                                             Predicted
                                                                             Predicted
           w/in each Subcomp.                        Obtained Model
            w/in each Subcomp.                                           Forest Condition
                                                                         Forest Condition

Copyright (C) Mitsubishi Research Institute, Inc.      11
Variety in a Subcompartment
                           There is a large variety inside a subcompartment
                        Non-forest area
                         • Deforestation area
                         • Canopy gaps
                        Invading woods other than planted species
                          • they’re not recorded on forest register


(Subcompartment)




Copyright (C) Mitsubishi Research Institute, Inc.   12
Experimental Setting (1/2)
Dataset:
   Target category: 4 species
               Larix kaempferi, Abies sachalinensis, Picea glehnii, other Broadleaf
      Source
               Hyperspectral reflectance by CASI-3
                      84 bands, 400 – 1060 nm

               9 signals given by PALSAR data
                      polarimetries (HH/HV/VV)
                      Three scattering components proposed by Freeman
                         i.e. surface scattering, double bounce scattering, volume scattering
                      Averaged alpha angle
                      Polarimetric entropy
                      Anisotropy

      Quadrats:




Copyright (C) Mitsubishi Research Institute, Inc.       13
Experimental Setting (2/2)
  Validation:
        100 times iteration of 5-fold cross valiadtion

  Comparison:
   Methodology
     Classification
              Spectral Angular Mapper ; SAM
              Regularized Discriminant Analysis; RDA (L2-norm regularization)
              ν-Support Vector Machines; SVM         (w/ Linear and RBF kernel)

        Regression
         Partial Least Squares; PLS

   Input data                                             #      wavelength range (nm)
      pseudo multi-spectral image                       Band 1        520 – 600
              ASTER image simulated from CASI-3 data    Band 2        630 – 690
              3 bands                                   Band 3        760 - 860
Copyright (C) Mitsubishi Research Institute, Inc.   14
Result: Quadrat Species Classification

      SDA achieves highest performance (87% prediction accuracy)
      Results by using hyperspectral images outperform pseudo-images

■ Prediction accuracy (Mean of 100 times 5-fold CV)
            Method                   SDA              SAM               RDA         SVM(Lin.)       SVM(RBF)
        CV-Accuracy                87.0 %            69.0 %            82.2 %         84.0           83.0 %

■ Confusion Matrix(SDA, Mean of 100 times 5-fold CV)
            Upper:Hyperspec                                                Predicted
            Lower: Multi-spec                       Larix              Abies           Picea         Broadleaf
                                                      92.0    %           1.8   %         6.2   %         0.0    %
                             Larix
                                                      78.8    %          16.3   %         4.3   %         0.8    %
                                                       0.4    %          84.4   %        15.1   %         0.1    %
                             Abies
                                                      23.3    %          76.4   %         0.3   %         0.0    %
        Actual
                                                      20.2    %           6.2   %        70.8   %         2.8    %
                             Picea
                                                      22.4    %           0.0   %        77.6   %         0.0    %
                                                       7.6    %           0.0   %         6.0   %        86.4    %
                          Braodleaf
                                                      32.7    %           0.2   %         0.0   %        67.1    %
Copyright (C) Mitsubishi Research Institute, Inc.                 15
Result: Quadrat Stand Factor Regression

      LASSO with hyperspectral data provides best performance for all stand factor

   RMSE
(10-fold CV)




 Prediction
(10-fold CV)                Canopy Cover              Timber Volume   Tree Height
                                                    poor results




Copyright (C) Mitsubishi Research Institute, Inc.     16
Result: Forest Extraction by Semi-supervised LDA
                                                                  Cloud
                       deforestation
                                                                  Bare ground
                                                                  Grass & Bamboo



                                                             ※ Dots indicate top-left
                                                               point of each pixel




                                                    Clouds


                        Sparse forest



Validation with 39 pixels selected manually
    Forest vs. non-forest:       100 %
    Overall Accuracy:           97.4 % (38/39)
Copyright (C) Mitsubishi Research Institute, Inc.    17
Result: Tree Species Composition in Subcompartments

    Confirm consistency between actual and estimated species by field survey




                                                                                   Legend

                                                                              Larix kaempferi

         Abies sachalinensis                        30 m resized pixels       Abies sachalinensis

                                                                              Picea glehnii

                                                                              Natural Broadleaf

                                                                          ▲ Field Survey Points

                                                                          ※ Dots indicate top-left
                                                                            point of each pixel


                            Broadleaf               Original CASI-3
                                    Invading broadleaf trees were found
Copyright (C) Mitsubishi Research Institute, Inc.           18
Predicted Tree Species Distribution Map
registry
                                                         Young Picea glehnii



                                                         Legend
Invasion of broadleaf
   to larch plantation


Predicted




                                                         mixed = below 70% dominancy

Copyright (C) Mitsubishi Research Institute, Inc.   19
Predicted Maps
       Canopy Cover                                 Timber Volume   Tree Height




Copyright (C) Mitsubishi Research Institute, Inc.         20
Validation and Evaluation – Canopy Cover Map

      Confirm prediction reflects forest conditions rightly by field survey




           non-thinned                                               non-thinned
                                              Line-thinned                         Line-thinned




                                                    Canopy Density




Copyright (C) Mitsubishi Research Institute, Inc.             21
Validation and Evaluation – Tree Height Map




                                                              Young Picea glehnii
                                                         & Broadleaf forest of low height




           Larix trees with relatively higher
               stand age were observed




Copyright (C) Mitsubishi Research Institute, Inc.   22
Conclusions

 Present forest monitoring method from hyperspectral and SAR image
     Integrate diverse data source with different property of targets
     To overcome high-dimension-small-sample-size problem resulting in over-fitting,
      sparse regularization techs (LASSO & Sparse Discriminant Analysis) are adopted


 3 advantages of sparse regularization
     Generalization, Interpretability, Optimal Band Selection


 Experimental simulations of satellite-based operation prove effectiveness
     Advantage in prediction accuracy to several supervised methods
     Advantage of hyperspectral data to multispectral
     Prediction results reflect existing forest conditions rightly



 Copyright (C) Mitsubishi Research Institute, Inc.   23
Many thanks for your kind attention.

                                                    Questions ?




Copyright (C) Mitsubishi Research Institute, Inc.       24
Supplementary Slides




Copyright (C) Mitsubishi Research Institute, Inc.   25
How sparse ?
          Canopy Cover Model: 8 / 84 params. = 9.5%               Tree Height Model: 14/ 93 params. = 15.0%

              #             Parameter               Coef.            #        Parameter          Coef.
              1                418.28         nm      6.32           1          473.31 nm          12.51
              2                441.81         nm     -1.36           2          560.97 nm         -22.81
              3                520.95         nm     -4.20
                                                                     3          640.38 nm          22.12
              4                626.01         nm     -1.07
                                                                     4          671.51 nm         -25.78
              5                766.86         nm     21.45
              6                923.70         nm      6.71           5          706.10 nm          18.61
              7                962.66         nm     -8.60           6          835.50 nm         -65.16
              8               1055.39         nm    -21.93           7          877.98 nm          27.13
              -              Intecept                75.98           8         1011.70 nm          40.04
          Timber Vol. Model: 5 / 93 params. = 5.4%                   9         1038.75 nm          42.92
              #           Parameter       Coef.                      10        1055.39 nm         -26.74
              1               426.17 nm    -0.59                     11        1061.01 nm         -31.73
              2               441.81 nm   -58.71                     12          HH                 -2.40
              3               513.13 nm   -18.67
                                                                     13           HV                -1.83
              4     Vol. Scattering Coef.   2.87
                                                                     14           VV                5.14
              5               VV           26.10
              -            Intercept      198.50                      -       Intercept            11.68
Copyright (C) Mitsubishi Research Institute, Inc.            26
Frequently Selected Parameters for Species Classification
 Distinct bands for confier/broadleaf properties & feature of species are selected.
      Around 450nm: absorption peak of G-type lignin richly contained in conifer wood
      Around 520nm: absorption peak of S-type lignin richly contained in broadleaf wood
      Red edge



                                                     出現率 Ratio [%]
                                                                     100
                                                                      90                                                                                     :freq. selected
   Frequency used                                                     80
                                                                      70
                                                                                                                                                              bands

                                              Appearance[%]
                                                                      60

 in prediction model                                                  50
                                                                      40
                                                                      30

by 100 bootstrapping                                                  20
                                                                      10
                                                                       0
                                                                           0   5   10   15   20   25   30   35    40 45   50   55   60   65   70   75   80

                                                                                                                 Band No.
                                                                                                                 バンド番号




                                                                                                                                                                  Picea
                                           Reflectance




        Reflectance                                                                                                                                               Larix
                                                                                                                                                                  Abies
          spectrum                                                                                                                                                Broadleaf


                                                                                                        Wavelength [nm]
 Copyright (C) Mitsubishi Research Institute, Inc.                                                27

More Related Content

What's hot

modulation transfer function (MTF)
modulation transfer function (MTF)modulation transfer function (MTF)
modulation transfer function (MTF)
AJAL A J
 
Noise figure limits for circular loop mr coils kumar john_hopkins
Noise figure limits for circular loop mr coils kumar john_hopkinsNoise figure limits for circular loop mr coils kumar john_hopkins
Noise figure limits for circular loop mr coils kumar john_hopkinsThiyagarajan K
 
Nityanand gopalika digital detectors for industrial applications
Nityanand gopalika   digital detectors for industrial applicationsNityanand gopalika   digital detectors for industrial applications
Nityanand gopalika digital detectors for industrial applications
Nityanand Gopalika
 
L16 radiation shielding
L16  radiation shieldingL16  radiation shielding
L16 radiation shielding
Mahbubul Hassan
 
L 15 radiation shielding principles. ppt
L 15  radiation shielding principles. pptL 15  radiation shielding principles. ppt
L 15 radiation shielding principles. ppt
Mahbubul Hassan
 
Squeezing Resoumis V1
Squeezing Resoumis V1Squeezing Resoumis V1
Squeezing Resoumis V1
Sidney Burks, Ph.D
 
Digital Radiography
Digital RadiographyDigital Radiography
Digital Radiography
Bipul Poudel
 
CR & DR
CR & DRCR & DR
Mg2521242127
Mg2521242127Mg2521242127
Mg2521242127
IJERA Editor
 
Image Quality - Radiologic Imaging
Image Quality - Radiologic ImagingImage Quality - Radiologic Imaging
Image Quality - Radiologic Imaging
Maria Nicole Sicaja
 
Study of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin Film
Study of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin FilmStudy of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin Film
Study of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin Film
IJERA Editor
 
Nityanand gopalika digital radiography performance study
Nityanand gopalika   digital radiography performance studyNityanand gopalika   digital radiography performance study
Nityanand gopalika digital radiography performance study
Nityanand Gopalika
 
Sensitometry3
Sensitometry3Sensitometry3
Sensitometry3mr_koky
 
Contrast
ContrastContrast
Contrastmr_koky
 

What's hot (17)

modulation transfer function (MTF)
modulation transfer function (MTF)modulation transfer function (MTF)
modulation transfer function (MTF)
 
Noise figure limits for circular loop mr coils kumar john_hopkins
Noise figure limits for circular loop mr coils kumar john_hopkinsNoise figure limits for circular loop mr coils kumar john_hopkins
Noise figure limits for circular loop mr coils kumar john_hopkins
 
Nityanand gopalika digital detectors for industrial applications
Nityanand gopalika   digital detectors for industrial applicationsNityanand gopalika   digital detectors for industrial applications
Nityanand gopalika digital detectors for industrial applications
 
L16 radiation shielding
L16  radiation shieldingL16  radiation shielding
L16 radiation shielding
 
L 15 radiation shielding principles. ppt
L 15  radiation shielding principles. pptL 15  radiation shielding principles. ppt
L 15 radiation shielding principles. ppt
 
Squeezing Resoumis V1
Squeezing Resoumis V1Squeezing Resoumis V1
Squeezing Resoumis V1
 
SPIE99_1B.DOC
SPIE99_1B.DOCSPIE99_1B.DOC
SPIE99_1B.DOC
 
Digital Radiography
Digital RadiographyDigital Radiography
Digital Radiography
 
CR & DR
CR & DRCR & DR
CR & DR
 
Mg2521242127
Mg2521242127Mg2521242127
Mg2521242127
 
Ijetcas14 479
Ijetcas14 479Ijetcas14 479
Ijetcas14 479
 
Image Quality - Radiologic Imaging
Image Quality - Radiologic ImagingImage Quality - Radiologic Imaging
Image Quality - Radiologic Imaging
 
Study of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin Film
Study of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin FilmStudy of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin Film
Study of Linear and Non-Linear Optical Parameters of Zinc Selenide Thin Film
 
Nityanand gopalika digital radiography performance study
Nityanand gopalika   digital radiography performance studyNityanand gopalika   digital radiography performance study
Nityanand gopalika digital radiography performance study
 
Cr & dr
Cr & drCr & dr
Cr & dr
 
Sensitometry3
Sensitometry3Sensitometry3
Sensitometry3
 
Contrast
ContrastContrast
Contrast
 

Viewers also liked

2_ullo_presentation.pdf
2_ullo_presentation.pdf2_ullo_presentation.pdf
2_ullo_presentation.pdfgrssieee
 
unrban-building-damage-detection-by-PJLi.ppt
unrban-building-damage-detection-by-PJLi.pptunrban-building-damage-detection-by-PJLi.ppt
unrban-building-damage-detection-by-PJLi.pptgrssieee
 
2_2011 IGARSS SMAP Applications Program Presentation.pdf
2_2011 IGARSS SMAP Applications Program Presentation.pdf2_2011 IGARSS SMAP Applications Program Presentation.pdf
2_2011 IGARSS SMAP Applications Program Presentation.pdfgrssieee
 
A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...
A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...
A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...grssieee
 
OHPIGARSS2011maeda1.pptx
OHPIGARSS2011maeda1.pptxOHPIGARSS2011maeda1.pptx
OHPIGARSS2011maeda1.pptxgrssieee
 
TH4_T0_04_thome.pptx
TH4_T0_04_thome.pptxTH4_T0_04_thome.pptx
TH4_T0_04_thome.pptxgrssieee
 

Viewers also liked (6)

2_ullo_presentation.pdf
2_ullo_presentation.pdf2_ullo_presentation.pdf
2_ullo_presentation.pdf
 
unrban-building-damage-detection-by-PJLi.ppt
unrban-building-damage-detection-by-PJLi.pptunrban-building-damage-detection-by-PJLi.ppt
unrban-building-damage-detection-by-PJLi.ppt
 
2_2011 IGARSS SMAP Applications Program Presentation.pdf
2_2011 IGARSS SMAP Applications Program Presentation.pdf2_2011 IGARSS SMAP Applications Program Presentation.pdf
2_2011 IGARSS SMAP Applications Program Presentation.pdf
 
A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...
A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...
A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE...
 
OHPIGARSS2011maeda1.pptx
OHPIGARSS2011maeda1.pptxOHPIGARSS2011maeda1.pptx
OHPIGARSS2011maeda1.pptx
 
TH4_T0_04_thome.pptx
TH4_T0_04_thome.pptxTH4_T0_04_thome.pptx
TH4_T0_04_thome.pptx
 

Similar to A_Methodology_of_Forest_Monitoring_from_Hyperspectral_Images_with_Sparse_Regularization1.pdf

Ijcet 06 09_001
Ijcet 06 09_001Ijcet 06 09_001
Ijcet 06 09_001
IAEME Publication
 
J Fernandes Hst 2009
J Fernandes Hst 2009J Fernandes Hst 2009
J Fernandes Hst 2009
jlfernandes85
 
Purkinje imaging for crystalline lens density measurement
Purkinje imaging for crystalline lens density measurementPurkinje imaging for crystalline lens density measurement
Purkinje imaging for crystalline lens density measurement
PetteriTeikariPhD
 
Ieee gold 2010 resta
Ieee gold 2010 restaIeee gold 2010 resta
Ieee gold 2010 restagrssieee
 
How to get bankable meteo data
How to get bankable meteo dataHow to get bankable meteo data
How to get bankable meteo data
Solar Reference
 
Phd final slides
Phd final slidesPhd final slides
Phd final slides
Andrea Barucci
 
Kps Environment 2 D3 D Wind Profiling
Kps Environment 2 D3 D Wind ProfilingKps Environment 2 D3 D Wind Profiling
Kps Environment 2 D3 D Wind Profiling
Sailreddragon
 
High-Speed Single-Photon SPAD Camera
High-Speed Single-Photon SPAD CameraHigh-Speed Single-Photon SPAD Camera
High-Speed Single-Photon SPAD Camera
Fabrizio Guerrieri
 
Sparse and Redundant Representations: Theory and Applications
Sparse and Redundant Representations: Theory and ApplicationsSparse and Redundant Representations: Theory and Applications
Sparse and Redundant Representations: Theory and Applications
Distinguished Lecturer Series - Leon The Mathematician
 
Noise removal techniques for microwave remote sensing radar data and its eval...
Noise removal techniques for microwave remote sensing radar data and its eval...Noise removal techniques for microwave remote sensing radar data and its eval...
Noise removal techniques for microwave remote sensing radar data and its eval...
csandit
 
NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...
NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...
NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...
cscpconf
 
Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...
Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...
Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...
The Air Force Office of Scientific Research
 
2
22
IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...
IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...
IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...
IRJET Journal
 
Super-Resolution of Multispectral Images
Super-Resolution of Multispectral ImagesSuper-Resolution of Multispectral Images
Super-Resolution of Multispectral Images
ijsrd.com
 
A REVIEW ON SYNTHETIC APERTURE RADAR
A REVIEW ON SYNTHETIC APERTURE RADARA REVIEW ON SYNTHETIC APERTURE RADAR
A REVIEW ON SYNTHETIC APERTURE RADAR
AM Publications
 
Presentation_Guccione.pptx
Presentation_Guccione.pptxPresentation_Guccione.pptx
Presentation_Guccione.pptxgrssieee
 
Defying Nyquist in Analog to Digital Conversion
Defying Nyquist in Analog to Digital ConversionDefying Nyquist in Analog to Digital Conversion
Defying Nyquist in Analog to Digital Conversion
Distinguished Lecturer Series - Leon The Mathematician
 
Fcv taxo chellappa
Fcv taxo chellappaFcv taxo chellappa
Fcv taxo chellappazukun
 

Similar to A_Methodology_of_Forest_Monitoring_from_Hyperspectral_Images_with_Sparse_Regularization1.pdf (20)

Ijcet 06 09_001
Ijcet 06 09_001Ijcet 06 09_001
Ijcet 06 09_001
 
J Fernandes Hst 2009
J Fernandes Hst 2009J Fernandes Hst 2009
J Fernandes Hst 2009
 
Purkinje imaging for crystalline lens density measurement
Purkinje imaging for crystalline lens density measurementPurkinje imaging for crystalline lens density measurement
Purkinje imaging for crystalline lens density measurement
 
Ieee gold 2010 resta
Ieee gold 2010 restaIeee gold 2010 resta
Ieee gold 2010 resta
 
How to get bankable meteo data
How to get bankable meteo dataHow to get bankable meteo data
How to get bankable meteo data
 
Phd final slides
Phd final slidesPhd final slides
Phd final slides
 
Kps Environment 2 D3 D Wind Profiling
Kps Environment 2 D3 D Wind ProfilingKps Environment 2 D3 D Wind Profiling
Kps Environment 2 D3 D Wind Profiling
 
High-Speed Single-Photon SPAD Camera
High-Speed Single-Photon SPAD CameraHigh-Speed Single-Photon SPAD Camera
High-Speed Single-Photon SPAD Camera
 
Sparse and Redundant Representations: Theory and Applications
Sparse and Redundant Representations: Theory and ApplicationsSparse and Redundant Representations: Theory and Applications
Sparse and Redundant Representations: Theory and Applications
 
Noise removal techniques for microwave remote sensing radar data and its eval...
Noise removal techniques for microwave remote sensing radar data and its eval...Noise removal techniques for microwave remote sensing radar data and its eval...
Noise removal techniques for microwave remote sensing radar data and its eval...
 
NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...
NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...
NOISE REMOVAL TECHNIQUES FOR MICROWAVE REMOTE SENSING RADAR DATA AND ITS EVAL...
 
Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...
Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...
Reinhardt - Adaptive Combinatorial Multimodal Sensing Physics & Methods - Spr...
 
2
22
2
 
Cm32546555
Cm32546555Cm32546555
Cm32546555
 
IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...
IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...
IRJET- Optimum Band Selection in Sentinel-2A Satellite for Crop Classificatio...
 
Super-Resolution of Multispectral Images
Super-Resolution of Multispectral ImagesSuper-Resolution of Multispectral Images
Super-Resolution of Multispectral Images
 
A REVIEW ON SYNTHETIC APERTURE RADAR
A REVIEW ON SYNTHETIC APERTURE RADARA REVIEW ON SYNTHETIC APERTURE RADAR
A REVIEW ON SYNTHETIC APERTURE RADAR
 
Presentation_Guccione.pptx
Presentation_Guccione.pptxPresentation_Guccione.pptx
Presentation_Guccione.pptx
 
Defying Nyquist in Analog to Digital Conversion
Defying Nyquist in Analog to Digital ConversionDefying Nyquist in Analog to Digital Conversion
Defying Nyquist in Analog to Digital Conversion
 
Fcv taxo chellappa
Fcv taxo chellappaFcv taxo chellappa
Fcv taxo chellappa
 

More from grssieee

Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...
Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...
Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...grssieee
 
SEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODEL
SEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODELSEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODEL
SEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODELgrssieee
 
TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...
TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...
TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...grssieee
 
THE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIES
THE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIESTHE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIES
THE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIESgrssieee
 
GMES SPACE COMPONENT:PROGRAMMATIC STATUS
GMES SPACE COMPONENT:PROGRAMMATIC STATUSGMES SPACE COMPONENT:PROGRAMMATIC STATUS
GMES SPACE COMPONENT:PROGRAMMATIC STATUSgrssieee
 
PROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETER
PROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETERPROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETER
PROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETERgrssieee
 
DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...
DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...
DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...grssieee
 
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...grssieee
 
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...grssieee
 
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...grssieee
 
test 34mb wo animations
test  34mb wo animationstest  34mb wo animations
test 34mb wo animationsgrssieee
 
2011_Fox_Tax_Worksheets.pdf
2011_Fox_Tax_Worksheets.pdf2011_Fox_Tax_Worksheets.pdf
2011_Fox_Tax_Worksheets.pdfgrssieee
 
DLR open house
DLR open houseDLR open house
DLR open housegrssieee
 
DLR open house
DLR open houseDLR open house
DLR open housegrssieee
 
DLR open house
DLR open houseDLR open house
DLR open housegrssieee
 
Tana_IGARSS2011.ppt
Tana_IGARSS2011.pptTana_IGARSS2011.ppt
Tana_IGARSS2011.pptgrssieee
 
Solaro_IGARSS_2011.ppt
Solaro_IGARSS_2011.pptSolaro_IGARSS_2011.ppt
Solaro_IGARSS_2011.pptgrssieee
 

More from grssieee (20)

Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...
Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...
Tangent height accuracy of Superconducting Submillimeter-Wave Limb-Emission S...
 
SEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODEL
SEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODELSEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODEL
SEGMENTATION OF POLARIMETRIC SAR DATA WITH A MULTI-TEXTURE PRODUCT MODEL
 
TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...
TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...
TWO-POINT STATISTIC OF POLARIMETRIC SAR DATA TWO-POINT STATISTIC OF POLARIMET...
 
THE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIES
THE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIESTHE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIES
THE SENTINEL-1 MISSION AND ITS APPLICATION CAPABILITIES
 
GMES SPACE COMPONENT:PROGRAMMATIC STATUS
GMES SPACE COMPONENT:PROGRAMMATIC STATUSGMES SPACE COMPONENT:PROGRAMMATIC STATUS
GMES SPACE COMPONENT:PROGRAMMATIC STATUS
 
PROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETER
PROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETERPROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETER
PROGRESSES OF DEVELOPMENT OF CFOSAT SCATTEROMETER
 
DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...
DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...
DEVELOPMENT OF ALGORITHMS AND PRODUCTS FOR SUPPORTING THE ITALIAN HYPERSPECTR...
 
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
 
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
 
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
EO-1/HYPERION: NEARING TWELVE YEARS OF SUCCESSFUL MISSION SCIENCE OPERATION A...
 
Test
TestTest
Test
 
test 34mb wo animations
test  34mb wo animationstest  34mb wo animations
test 34mb wo animations
 
Test 70MB
Test 70MBTest 70MB
Test 70MB
 
Test 70MB
Test 70MBTest 70MB
Test 70MB
 
2011_Fox_Tax_Worksheets.pdf
2011_Fox_Tax_Worksheets.pdf2011_Fox_Tax_Worksheets.pdf
2011_Fox_Tax_Worksheets.pdf
 
DLR open house
DLR open houseDLR open house
DLR open house
 
DLR open house
DLR open houseDLR open house
DLR open house
 
DLR open house
DLR open houseDLR open house
DLR open house
 
Tana_IGARSS2011.ppt
Tana_IGARSS2011.pptTana_IGARSS2011.ppt
Tana_IGARSS2011.ppt
 
Solaro_IGARSS_2011.ppt
Solaro_IGARSS_2011.pptSolaro_IGARSS_2011.ppt
Solaro_IGARSS_2011.ppt
 

Recently uploaded

Introduction to CHERI technology - Cybersecurity
Introduction to CHERI technology - CybersecurityIntroduction to CHERI technology - Cybersecurity
Introduction to CHERI technology - Cybersecurity
mikeeftimakis1
 
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdfFIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance
 
Enhancing Performance with Globus and the Science DMZ
Enhancing Performance with Globus and the Science DMZEnhancing Performance with Globus and the Science DMZ
Enhancing Performance with Globus and the Science DMZ
Globus
 
PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)
Ralf Eggert
 
SAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdf
SAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdfSAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdf
SAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdf
Peter Spielvogel
 
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Albert Hoitingh
 
A tale of scale & speed: How the US Navy is enabling software delivery from l...
A tale of scale & speed: How the US Navy is enabling software delivery from l...A tale of scale & speed: How the US Navy is enabling software delivery from l...
A tale of scale & speed: How the US Navy is enabling software delivery from l...
sonjaschweigert1
 
RESUME BUILDER APPLICATION Project for students
RESUME BUILDER APPLICATION Project for studentsRESUME BUILDER APPLICATION Project for students
RESUME BUILDER APPLICATION Project for students
KAMESHS29
 
FIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdf
FIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdfFIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdf
FIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdf
FIDO Alliance
 
The Metaverse and AI: how can decision-makers harness the Metaverse for their...
The Metaverse and AI: how can decision-makers harness the Metaverse for their...The Metaverse and AI: how can decision-makers harness the Metaverse for their...
The Metaverse and AI: how can decision-makers harness the Metaverse for their...
Jen Stirrup
 
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
DanBrown980551
 
Accelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish CachingAccelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish Caching
Thijs Feryn
 
Monitoring Java Application Security with JDK Tools and JFR Events
Monitoring Java Application Security with JDK Tools and JFR EventsMonitoring Java Application Security with JDK Tools and JFR Events
Monitoring Java Application Security with JDK Tools and JFR Events
Ana-Maria Mihalceanu
 
PCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase TeamPCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase Team
ControlCase
 
Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?
Nexer Digital
 
Generative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to ProductionGenerative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to Production
Aggregage
 
Essentials of Automations: The Art of Triggers and Actions in FME
Essentials of Automations: The Art of Triggers and Actions in FMEEssentials of Automations: The Art of Triggers and Actions in FME
Essentials of Automations: The Art of Triggers and Actions in FME
Safe Software
 
Removing Uninteresting Bytes in Software Fuzzing
Removing Uninteresting Bytes in Software FuzzingRemoving Uninteresting Bytes in Software Fuzzing
Removing Uninteresting Bytes in Software Fuzzing
Aftab Hussain
 
Free Complete Python - A step towards Data Science
Free Complete Python - A step towards Data ScienceFree Complete Python - A step towards Data Science
Free Complete Python - A step towards Data Science
RinaMondal9
 
Securing your Kubernetes cluster_ a step-by-step guide to success !
Securing your Kubernetes cluster_ a step-by-step guide to success !Securing your Kubernetes cluster_ a step-by-step guide to success !
Securing your Kubernetes cluster_ a step-by-step guide to success !
KatiaHIMEUR1
 

Recently uploaded (20)

Introduction to CHERI technology - Cybersecurity
Introduction to CHERI technology - CybersecurityIntroduction to CHERI technology - Cybersecurity
Introduction to CHERI technology - Cybersecurity
 
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdfFIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
 
Enhancing Performance with Globus and the Science DMZ
Enhancing Performance with Globus and the Science DMZEnhancing Performance with Globus and the Science DMZ
Enhancing Performance with Globus and the Science DMZ
 
PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)
 
SAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdf
SAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdfSAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdf
SAP Sapphire 2024 - ASUG301 building better apps with SAP Fiori.pdf
 
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
 
A tale of scale & speed: How the US Navy is enabling software delivery from l...
A tale of scale & speed: How the US Navy is enabling software delivery from l...A tale of scale & speed: How the US Navy is enabling software delivery from l...
A tale of scale & speed: How the US Navy is enabling software delivery from l...
 
RESUME BUILDER APPLICATION Project for students
RESUME BUILDER APPLICATION Project for studentsRESUME BUILDER APPLICATION Project for students
RESUME BUILDER APPLICATION Project for students
 
FIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdf
FIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdfFIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdf
FIDO Alliance Osaka Seminar: Passkeys and the Road Ahead.pdf
 
The Metaverse and AI: how can decision-makers harness the Metaverse for their...
The Metaverse and AI: how can decision-makers harness the Metaverse for their...The Metaverse and AI: how can decision-makers harness the Metaverse for their...
The Metaverse and AI: how can decision-makers harness the Metaverse for their...
 
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
 
Accelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish CachingAccelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish Caching
 
Monitoring Java Application Security with JDK Tools and JFR Events
Monitoring Java Application Security with JDK Tools and JFR EventsMonitoring Java Application Security with JDK Tools and JFR Events
Monitoring Java Application Security with JDK Tools and JFR Events
 
PCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase TeamPCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase Team
 
Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?
 
Generative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to ProductionGenerative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to Production
 
Essentials of Automations: The Art of Triggers and Actions in FME
Essentials of Automations: The Art of Triggers and Actions in FMEEssentials of Automations: The Art of Triggers and Actions in FME
Essentials of Automations: The Art of Triggers and Actions in FME
 
Removing Uninteresting Bytes in Software Fuzzing
Removing Uninteresting Bytes in Software FuzzingRemoving Uninteresting Bytes in Software Fuzzing
Removing Uninteresting Bytes in Software Fuzzing
 
Free Complete Python - A step towards Data Science
Free Complete Python - A step towards Data ScienceFree Complete Python - A step towards Data Science
Free Complete Python - A step towards Data Science
 
Securing your Kubernetes cluster_ a step-by-step guide to success !
Securing your Kubernetes cluster_ a step-by-step guide to success !Securing your Kubernetes cluster_ a step-by-step guide to success !
Securing your Kubernetes cluster_ a step-by-step guide to success !
 

A_Methodology_of_Forest_Monitoring_from_Hyperspectral_Images_with_Sparse_Regularization1.pdf

  • 1. IGARSS 2011, Vancouver TU4.T08.1: Hyperspectral Monitoring of the Environment I A Methodology of Forest Monitoring from Hyperspectral Images with Sparse Regularization Jul. 26, 2011 MITSUBISHI RESEARCH INSTITUTE, INC. Keigo YOSHIDA, Takashi OHKI, Masahiro TERABE, Hozuma SEKINE (MRI) Tomomi TAKEDA (ERSDAC) Copyright (C) Mitsubishi Research Institute, Inc.
  • 2. Introduction:Forest Monitoring by Remote Sensing  Decision making for Forest Management  Disaster prevention planning Accurate Info. Needs  Finding forest of poor growth Needs of Present Forest  GHG Credit estimation  Resource management  Forest conditions change dynamically High Cost of  Conduct periodical field survey Problem Forest Survey Problem  estimate 600-1200 USD / Km2 / year & Monitoring ※ in case of Japan; 1 USD = 81 yen Forest Monitoring  No need for field survey all over the area Solution Solution by Remote Sensing  Highly-frequent observation Copyright (C) Mitsubishi Research Institute, Inc. 2
  • 3. Introduction:Highly-developed Sensing Tech. & Challenges Sensing Tech. Challenges  Hard to bring out potential of big sensor data Hyperspectral sensor [e.g.] NDVI use just 2 bands, or Red and IR & have to select optimal band combinations  provide detailed optical info. on forest physiognomy  Complexity of prediction model increases,  growth situation resulting in poor prediction performance  character of tree species  Dimension is high but sample size is small etc. due to limitation of field survey  This causes model overfitting Sensor fusion  Modeling is not easy for several sensor data  Reflect diverse property of different physical property of targets  Statistical or Data-driven approach is needed Copyright (C) Mitsubishi Research Institute, Inc. 3
  • 4. Research Outline Utilize rich data by a machine learning technique (sparse regularization) and achieve accurate, informative, & less costly forest monitoring  Remote Sensing Data Fusion Input Data (CASI-3 hyperspectral images + SAR signals) Input Data  Field Survey Results  Sparse Regularization Methodology Methodology (Sparse Discriminant Analysis、LASSO regression)  Predicted Stand Factors of each subcompartments for Forest Management Output Data Output Data (Species, Canopy cover, Timber volume, Tree height)  Prediction Models Subcompartment: a general spatial unit for forest monitoring Copyright (C) Mitsubishi Research Institute, Inc. 4
  • 5. Target Site Town-owned forest in Shimokawa, Hokkaido, Japan  Approx. 90 % of town is covered by forest  Utilize local conifer resources for business  Environmental model city for low-carbon society Shimokawa Copyright (C) Mitsubishi Research Institute, Inc. 5
  • 6. Remote Sensing Data Species and Canopy-cover prediction: CASI-3 hyperspectrum Volume and Height prediction: Data fusion (CASI-3 + PALSAR)  Remote Sensing  Hyperspectral sensor (optical property)  Airborne hyperspectral imager CASI-3  84 bands from 400 to 1060 nm (wavelenght res. : 8 nm)  Original spatial res.: 2.0 m → Resolution is decreased to 30m to simulate satellite-based operation  PALSAR (shape or volume property)  Microwave backscattering org. image resized Copyright (C) Mitsubishi Research Institute, Inc. 6
  • 7. Field Survey During aircraft obs., conduct field survey to collect data for modeling & validation  Field Survey:  Place 25-sq-m quadrats  Inventory study for trees whose DBH > 5cm & and record tree species  Canopy cover measurement with whole-sky camera  Height measurement for sampled 10 trees Copyright (C) Mitsubishi Research Institute, Inc. 7
  • 8. What is Sparse Regularization ? Why Do I Use it ? “Sparse” means the model has a low # of nonzero parameters ■ Optimal Band Selection Ineffective parameters will be removed from prediction model automatically by solving convex optimization problem ■ Higher Generalization Capability simple model with smaller # of bands achieves less overfitting; better prediction performance ■ More Interpretable Model Copyright (C) Mitsubishi Research Institute, Inc. 8
  • 9. Sparse Regularization:Theoretical Overview Add penalty to loss function to obtain model with small num. of variables  LASSO (R. Tibshirani et al., 96) Loss function (LS) (penalty) norm  Optimal Scoring (T. Hastie et al., 94)  Perform Fisher’s linear discriminant analysis as regression by score  convert categorical variables for class membership into quantitative  Optimize and weight vector simultaneously Copyright (C) Mitsubishi Research Institute, Inc. 9
  • 10. Intuitive Explanation of Sparse Regularization  To reduce empirical errors, <penalty> W moves away from 0, then penalty increases  L1-norm: attraction force to 0 is const. -> Small values in W tend to be 0 <attraction force to 0> L2-norm: attraction force is small around 0 -> Small values in W remain  Coefficients L1-regularization L2-regularization Copyright (C) Mitsubishi Research Institute, Inc. 10
  • 11. Experimental Flow 1. Modeling Prediction Prediction Performance Performance Hyperspectral Reflectance Hyperspectral Reflectance Classification Classification Sparse LDA (ave. w/in each quadrat) (ave. w/in each quadrat) Model Model Regression Regression PALSAR Signals PALSAR Signals LASSO Regression Model Model 2. Prediction for Subcompartments Forest Pixel Extraction Hyperspectral Reflectance Hyperspectral Reflectance Semisupervised LDA Forest Pixels Forest Pixels (30m x 30m pixels) (30m x 30m pixels) Subcompartment Prediction Averaged Reflectance Averaged Reflectance Predicted Predicted w/in each Subcomp. Obtained Model w/in each Subcomp. Forest Condition Forest Condition Copyright (C) Mitsubishi Research Institute, Inc. 11
  • 12. Variety in a Subcompartment There is a large variety inside a subcompartment Non-forest area • Deforestation area • Canopy gaps Invading woods other than planted species • they’re not recorded on forest register (Subcompartment) Copyright (C) Mitsubishi Research Institute, Inc. 12
  • 13. Experimental Setting (1/2) Dataset:  Target category: 4 species  Larix kaempferi, Abies sachalinensis, Picea glehnii, other Broadleaf  Source  Hyperspectral reflectance by CASI-3  84 bands, 400 – 1060 nm  9 signals given by PALSAR data  polarimetries (HH/HV/VV)  Three scattering components proposed by Freeman i.e. surface scattering, double bounce scattering, volume scattering  Averaged alpha angle  Polarimetric entropy  Anisotropy  Quadrats: Copyright (C) Mitsubishi Research Institute, Inc. 13
  • 14. Experimental Setting (2/2) Validation:  100 times iteration of 5-fold cross valiadtion Comparison:  Methodology  Classification  Spectral Angular Mapper ; SAM  Regularized Discriminant Analysis; RDA (L2-norm regularization)  ν-Support Vector Machines; SVM (w/ Linear and RBF kernel)  Regression Partial Least Squares; PLS  Input data # wavelength range (nm)  pseudo multi-spectral image Band 1 520 – 600  ASTER image simulated from CASI-3 data Band 2 630 – 690  3 bands Band 3 760 - 860 Copyright (C) Mitsubishi Research Institute, Inc. 14
  • 15. Result: Quadrat Species Classification SDA achieves highest performance (87% prediction accuracy) Results by using hyperspectral images outperform pseudo-images ■ Prediction accuracy (Mean of 100 times 5-fold CV) Method SDA SAM RDA SVM(Lin.) SVM(RBF) CV-Accuracy 87.0 % 69.0 % 82.2 % 84.0 83.0 % ■ Confusion Matrix(SDA, Mean of 100 times 5-fold CV) Upper:Hyperspec Predicted Lower: Multi-spec Larix Abies Picea Broadleaf 92.0 % 1.8 % 6.2 % 0.0 % Larix 78.8 % 16.3 % 4.3 % 0.8 % 0.4 % 84.4 % 15.1 % 0.1 % Abies 23.3 % 76.4 % 0.3 % 0.0 % Actual 20.2 % 6.2 % 70.8 % 2.8 % Picea 22.4 % 0.0 % 77.6 % 0.0 % 7.6 % 0.0 % 6.0 % 86.4 % Braodleaf 32.7 % 0.2 % 0.0 % 67.1 % Copyright (C) Mitsubishi Research Institute, Inc. 15
  • 16. Result: Quadrat Stand Factor Regression LASSO with hyperspectral data provides best performance for all stand factor RMSE (10-fold CV) Prediction (10-fold CV) Canopy Cover Timber Volume Tree Height poor results Copyright (C) Mitsubishi Research Institute, Inc. 16
  • 17. Result: Forest Extraction by Semi-supervised LDA Cloud deforestation Bare ground Grass & Bamboo ※ Dots indicate top-left point of each pixel Clouds Sparse forest Validation with 39 pixels selected manually  Forest vs. non-forest: 100 %  Overall Accuracy: 97.4 % (38/39) Copyright (C) Mitsubishi Research Institute, Inc. 17
  • 18. Result: Tree Species Composition in Subcompartments Confirm consistency between actual and estimated species by field survey Legend Larix kaempferi Abies sachalinensis 30 m resized pixels Abies sachalinensis Picea glehnii Natural Broadleaf ▲ Field Survey Points ※ Dots indicate top-left point of each pixel Broadleaf Original CASI-3 Invading broadleaf trees were found Copyright (C) Mitsubishi Research Institute, Inc. 18
  • 19. Predicted Tree Species Distribution Map registry Young Picea glehnii Legend Invasion of broadleaf to larch plantation Predicted mixed = below 70% dominancy Copyright (C) Mitsubishi Research Institute, Inc. 19
  • 20. Predicted Maps Canopy Cover Timber Volume Tree Height Copyright (C) Mitsubishi Research Institute, Inc. 20
  • 21. Validation and Evaluation – Canopy Cover Map Confirm prediction reflects forest conditions rightly by field survey non-thinned non-thinned Line-thinned Line-thinned Canopy Density Copyright (C) Mitsubishi Research Institute, Inc. 21
  • 22. Validation and Evaluation – Tree Height Map Young Picea glehnii & Broadleaf forest of low height Larix trees with relatively higher stand age were observed Copyright (C) Mitsubishi Research Institute, Inc. 22
  • 23. Conclusions  Present forest monitoring method from hyperspectral and SAR image  Integrate diverse data source with different property of targets  To overcome high-dimension-small-sample-size problem resulting in over-fitting, sparse regularization techs (LASSO & Sparse Discriminant Analysis) are adopted  3 advantages of sparse regularization  Generalization, Interpretability, Optimal Band Selection  Experimental simulations of satellite-based operation prove effectiveness  Advantage in prediction accuracy to several supervised methods  Advantage of hyperspectral data to multispectral  Prediction results reflect existing forest conditions rightly Copyright (C) Mitsubishi Research Institute, Inc. 23
  • 24. Many thanks for your kind attention. Questions ? Copyright (C) Mitsubishi Research Institute, Inc. 24
  • 25. Supplementary Slides Copyright (C) Mitsubishi Research Institute, Inc. 25
  • 26. How sparse ? Canopy Cover Model: 8 / 84 params. = 9.5% Tree Height Model: 14/ 93 params. = 15.0% # Parameter Coef. # Parameter Coef. 1 418.28 nm 6.32 1 473.31 nm 12.51 2 441.81 nm -1.36 2 560.97 nm -22.81 3 520.95 nm -4.20 3 640.38 nm 22.12 4 626.01 nm -1.07 4 671.51 nm -25.78 5 766.86 nm 21.45 6 923.70 nm 6.71 5 706.10 nm 18.61 7 962.66 nm -8.60 6 835.50 nm -65.16 8 1055.39 nm -21.93 7 877.98 nm 27.13 - Intecept 75.98 8 1011.70 nm 40.04 Timber Vol. Model: 5 / 93 params. = 5.4% 9 1038.75 nm 42.92 # Parameter Coef. 10 1055.39 nm -26.74 1 426.17 nm -0.59 11 1061.01 nm -31.73 2 441.81 nm -58.71 12 HH -2.40 3 513.13 nm -18.67 13 HV -1.83 4 Vol. Scattering Coef. 2.87 14 VV 5.14 5 VV 26.10 - Intercept 198.50 - Intercept 11.68 Copyright (C) Mitsubishi Research Institute, Inc. 26
  • 27. Frequently Selected Parameters for Species Classification Distinct bands for confier/broadleaf properties & feature of species are selected.  Around 450nm: absorption peak of G-type lignin richly contained in conifer wood  Around 520nm: absorption peak of S-type lignin richly contained in broadleaf wood  Red edge 出現率 Ratio [%] 100 90 :freq. selected Frequency used 80 70 bands Appearance[%] 60 in prediction model 50 40 30 by 100 bootstrapping 20 10 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 Band No. バンド番号 Picea Reflectance Reflectance Larix Abies spectrum Broadleaf Wavelength [nm] Copyright (C) Mitsubishi Research Institute, Inc. 27