Docstoc

Damage Detection in Bourmedes City Algeria by Visual Analysis

Document Sample
Damage Detection in Bourmedes City Algeria by Visual Analysis Powered By Docstoc
					International Symposium on Stochastic Analyses
for Risk Management (SARM2010)




  Remote Sensing Technology for Urban
   Monitoring and Damage Assessment

                      December 23, 2010

                    Fumio YAMAZAKI
      Graduate School of Engineering, Chiba University, Japan.

                                                                 1
    Prof. Shinozuka and his associates
          at Columbia University




September 1984          August 1986
                                         2
Year     Prof. Masanobu Shinozuka      Affiliation             Research topics

1930   1930.12.23 born in Tokyo
1940
1950   1953 BS, Kyoto University       Kyoto         reliability theory
       1955 MS, Kyoto University       Columbia
1960   1960 PhD, Columbia University   Columbia      Monte Carlo simulation
       1969 Professor
1970                                   Columbia      system identification/control
                                                     stochastic dynamics
1980                                   Columbia      lifeline
       1988 Princeton University       Princeton     stochastic FEM
1990   1990 Director, NCEER            Princeton     earthquake engineering
       1995 USC                        USC           advanced technologies (including
                                                     remote sensing and GIS)
2000   2001 UC Irvine                  USC           risk management
                                       UC Irvine     monitoring of infrastructures
2010   2010. 12.23 80th Birthday

                                                                                        3
        Remote Sensing and Prof. Shinozuka
2000 MCEER-EDM Workshop      Joint Survey by MCEER-EDM
@ Newport Beach              after the 1999 Kocaeli, Turkey EQ




2003 1st Remote Sensing WS @UCI
                                  SAR Simulation of Urban Areas
                                  (Shinozuka et al. 2000)




                                                                  4
                   Contents
Basic issues of remote sensing
New high-resolution optical sensors
Thermal sensor and Lidar
Synthetic Aperture Radar (SAR)




                                       5
  Use of Remote Sensing in Disaster Management
Pre-event inventory and topography mapping:
 land-cover classification, building and road detection,
 DSM/DEM, temperature


Post-event damage assessment:
  change detection due to natural
 and man-made disasters


Combined use of texture and DEM is important
for pre- and post-event damage assessment.




                            Khao Lak, Thailand             6
 Sichuan, China
           Platforms of Remote Sensing
Space-borne                                        Space Shuttle
                Satellite
                Optical Sensor/SAR
                                 700-900km   185-575km

                Aerial Photography                             Airborne SAR
Airborne

                     1.2-3.5km
                                                                   10-12km
 Aerial Television

        0.3km



Ground-based


                                                                             7
Wavelength of Electromagnetic Waves and
            Satellite Sensors

                  Reflection    Radiation       Active
  g-ray, x-ray   UV       Infrared                Microwave
                  Visible                   Infrared

                                                       Wavelength




            B    G          R           NIR            Mid-IR Thermal



                                                                    8
   Color Composite Example of Landsat TM image

                                  (R, G, B)


                                 True color
                                 (3, 2, 1)




Natural color                    False color
(3, 4, 2)                        (4, 3, 2)



                                 Principles of Remote
                                                        9
                                 Sensing, ITC
   Near-infrared band can tell the difference of
           natural and manmade lawn              NIR  R
 QuickBird     True color                      NDVI 
                            False color                 NIR  R
Chiba Marine
Stadium
 Manmade
 lawn

                                          -1            +1
                                                             NDVI

Makuhari
Seaside Park

 Natural
 lawn
                                                                  10
              Spectral Reflectance of Earth-Surface Materials
                    and Spectral bands of Landsat TM
                           Typical S pectral Reflectance Curves
                                                                                Water (clear)
              80%
                                                                                Vegetation (green)
                          Bands of Landsat TM
              70%                                                               Dry bare soil (gray-brown)

                      Blue Green Red Near IR                         Mid IR                     Mid IR
              60%
                      1 2         3     4                                  5                          7 TM band
              50%
Reflectance




              40%                                                                        Soil
              30%                                                                        Vegetation
              20%

              10%
                                      Water
              0%
                    0.4     0.6       0.8      1.0   1.2      1.4    1.6       1.8    2.0       2.2      2.4   2.6
                                                           Wavelength ((mm)
                                                                       m m)
                                                                                                                 11
       Reflectance of vegetation measured on 2005/2/20
Reflectance=
(Radiation of an object)/ (Radiation of white reference)

       Kata Beach, Phuket, Thailand
       affected by 2006.12.26 Indian Ocean Tsunami
                   100
                                  B     G         R         NIR
                                                                                          Lawn planted
                   80                                                                     after tsunami
 Reflectance (%)




                   60                                                                     Grass
                                                                                          weakened
                   40                                                                     by tsunami
                                                                        grass 49
                                                                        yellow58
                   20                                                   dead 59           Dead lawn due
                                                                                          to tsunami
                     0
                      300   400   500       600       700   800   900   1000       1100
                                                                                                          12
                                      Wavelength (nm)
                   Contents
Basic issues of remote sensing
New high-resolution optical sensors
Thermal sensor and Lidar
Synthetic Aperture Radar (SAR)




                                       13
             QuickBird
Launch
on October 18, 2001

Features
Highest resolution sensors
commercially available

Sensor Resolution
61cm panchromatic at nadir
2.44 m multi-spectral at nadir
  Blue -Band 1 450-520 nm
  Green -Band 2 520-600 nm
  Red -Band 3   630-690 nm
  Near IR-Band 4 760-900 nm
  Panchromatic 450-900 nm
                                                               14
                                 http://www.digitalglobe.com
 Comparison of ASTER and QuickBird Images of Bam




ASTER 2003/10/28              QuickBird 2003/09/30
False color (B:G,G:R,R:NIR)   False color (B:G,G:R,R:NIR)
Pixel size : 15m              Pixel size : 0.6m




                                                            15
Damage classification of masonry buildings (EMS, 1998)
     and typical pre- and post event QB images
                                                                             EMS: European Macroseismic Scale
  Classification of damage to masonry buildings                                   Pre- event   Post- event
                Grade 1: Negligible to slight damage
                (no structural damage, slight non-
                structural damage)
                Hair-line cracks in very few walls.
                Fall of small pieces of plaster only.
                Fall of loose stones from upper parts of buildings in
                very few cases.
                Grade 2: Moderate damage
                (slight structural damage, moderate non-                                         Grade 3
                structural damage)
                Cracks in many walls.
                Fall of fairly large pieces of plaster.
                Partial collapse of chimneys.
                Grade 3: Substantial to heavy damage
                (moderate structural damage, heavy non-
                structural damage)
                Large and extensive cracks in most walls.
                Roof tiles detach. Chimneys fracture at the roof line;
                failure of individual non-structural elements (partitions,
                gable walls).
                                                                                                 Grade 4
                Grade 4: Very heavy damage
                (heavy structural damage, very heavy non-
                structural damage)
                Serious failure of walls; partial structural failure of
                roofs and floors.

                Grade 5: Destruction
                (very heavy structural damage)
                Total or near total collapse.
                                                                                                                16
                                                                                                 Grade 5
Building damage in Boumerdes City, Algeria
       by field survey and QB images




                                             17
   Correction of cast-shadows in RS images




       ■■■
       Shadow and non-shadow pairs
Original image               Corrected image




                                               18
    Original and corrected shadow-free QB images
                         False Color




        Original image             Corrected shadow-free image
                         Unsupervised
                         Classification




10 classes                      8 classes                        19
                           Digital Aerial Camera
4 Panchromatic
lenses                                 UltraCam D                                        DMC

                                 4 Multispectral
                                 lenses




 Geographical Survey Institute, Tsukuba, Japan

 http://www.vexcel.com/products/photogram/
 ultracam/index.html                               http://www.intergraph.com/dmc/default.asp   20
Kashiwazaki,
Niigata Pref.
July 2007
By Asia Air
Survey Co.




                21
                Object-based Classification
・For high-resolution images, pixel-based classification may cause salt-
and-pepper noises. To solve this, object-based classification is needed.

・In object-based classification, some pixels are converted into one object
considering spatial characteristics by image segmentation.
 The classification is conducted with respect to the object.



                   Object



                    Pixel



                    Concept of ‘pixel’ and ‘object’
                                                                           22
     Training data of the post-event digital aerial image
for supervised classification by maximum likelihood method




 Ground                                           Debris




   Tree                                          Red Roof




White Roof                            Grass                 23
              Road      Black Roof              Gray Roof
 Comparison of the results of object-based and pixel-based
         classifications for the post-event image




Pixel-based (Maximum Likelihood)                  Object-based
    Black Roof     White Roof
    Gray Roof      Red Roof        Parameters of the object-based method
    Road           Ground          Scale Parameter = 40
                   Grass
                                     Layer Weight = 1.0, Shape Factor = 0.5
    Vegetation
                                    Compact Weight = 0, Smooth Weight = 1.0
    Debris
                                                                              24
                   Contents
Basic issues of remote sensing
New high-resolution optical sensors
Thermal sensor and Lidar
Synthetic Aperture Radar (SAR)




                                       25
                                             Airborne Thermal
                                             Sensing of Tokyo


31.3℃                  28.1℃

                                             Date: August 7, 2007
                                             Time: day (13:00) & night (22:00)
                                             Weather: partly cloudy
(a) TD at 13:00 pm      (b) TN at 22:00 pm   Temperature:
                                              Max 33.2 deg C, Min 25.8 deg C.




                                             Termo Tracer TS7302
D=3.2℃                                       NEC Avio Infrared Technologies
                                             Wavelength: 8 mm - 14 mm
                                             Coverage temperature:
                                                    -40 deg. C to 120 deg. C
                                                                               26
(c) Difference TD-TN    (d) RGB color
Comparison of visible and thermal aerial images
                UltraCamD                    2006. 8. 7 pm 13:25
           Jingu Baseball Park: manmade lawn       TABI 320




 (a) True color        (b) False color                    28℃   53℃
                       composite (RGB=432) (c) Temperature
 composite (RGB=321)




                National Stadium: natural lawn                     27
Thermogram by Handy Thermal Imager
               Rooftop of Chiba U.   Termo Tracer TH910

                2008.9.2 13:30
                  30.3℃




                                        13:30   19:00   28
   Scanning Airborne Laser (LIDAR)

    GPS                               GPS
    Satellite                         Satellite



                           Laser
                           Altimeter
                         Foot Print (x, y, z)
                                      coordinate
        Measurement         GPS
        accuracy: 15cm                  Asia Air Survey Co.



LIDAR:
Light Detection
and Ranging                                               29
             LIDAR based change detection of
            dense urban areas (Roppongi, Tokyo)
                              Digital Surface Model by
                              LIDAR surveying flights




June 1999




              February 2004                              30
                                                         30
                   Contents
Basic issues of remote sensing
New high-resolution optical sensors
Thermal sensor and Lidar
Synthetic Aperture Radar (SAR)




                                       31
SAR: Synthetic Aperture Radar
      Active Microwave Sensor
      Emitting microwave signals,
      then receiving their reflection
      from objects on earth’s surface

       All Weather, Day and Nighttime




                  ERS/SAR
                  Wave Length: 5.7cm (C-band VV)
                  Resolution: 30m
                  Recurrent Period: 35 days
                                                   32
  Change Detection from SAR intensity images
 1.Image matching
 2.Speckle noise filtering (Lee Filter)
 3.Calculating following indices:
         Difference of backscattering
        coefficient (after – before)
      d [dB]  10  log 10 Iai  10  log 10 Ibi

         Correlation coefficient
                      N           N      N
                   N  Iai Ibi   Iai  Ibi
   r                 i 1        i 1   i 1

          N         N      N
                             2
                                                N    
                                                       2

                                               
         N  Iai   Iai   N  Ibi   Ibi  
                  2                         2
                   i 1    i 1      i 1  
          i 1                                     

Iai and Ibi are the digital numbers of the post- and pre-images.
 Īai and Ībi are the corresponding averaged digital numbers over the pixel window.   33
Estimated Damage Areas in 1995 Kobe EQ using ERS/SAR
  Difference of backscattering                                                                              Correlation coefficient
 coefficient (1995/5/23 – 1994/6/3)




                                                  1
   Correlation Coefficient of Intensity Images




                                                                         Filtering Window: 21 x 21
                                                                         Calculation Window: 13 x 13
                                                                                       95/05-94/10
                                                                                       95/05-94/06
                                                                                       95/05-93/08
                                                                                       95/05-92/11
                                                 0.5
                                                              Severe
                                                              damage area




                                                  0        Very severe
                                                           damage area
                                                   -3         -2        -1       0        1            2   Red: Very severe damage area
                                                       Difference of Backscattering Coefficient [dB]       Yellow: Severe damage area     34
 Color composite of the two TerraSAR-X images




                Tokyo     Toyosu         ■May 24th, 2008
Congressional station   development      ■Nov. 24th, 2009
office buildings            area
                                      • Red color shows the
                                        decrease of backscatter,
                                        e.g. removed buildings.

                                      • Cyan color shows the
                                        increase of backscatter,
                                        e.g. new buildings.

                                                               35
                Difference (d) and Correlation (r)
                                                            N         N    N
                                                         N  Ia  Ib   Ia  Ib
           d  Ia  Ib                      r              i          i    i

                                                  N 2       N
                                                                   2 
                                                                         N       N
                                                                                         
                                                  N  Ia  ( Ia )  N  Ib  ( Ib) 2 
                                                                             2


 Īa,b: The mean of σ0 within the window           i         i        i        i       
 a: May 24th, 2008; b: Nov. 24th, 2009       Ia,b: σ0 of ith pixel in the widow
Difference: d                             Correlation coefficient: r




                                                                                             36
Change detection based on d and r
      Histogram of the difference

        d                            ■Newly built
                                     ■Removed
                                     ■Low correlation




          μd =-0.94, σd =2.44
                                                                  Combined factor, z
Histogram of the correlation coefficient

        r                                          d
                                           z              cr
                                                Max( d )




            μr =0.60 , σr = 0.38
                                                                                  37
                 Upper House new office buildings
                                   (a) Spring in 2009




     Range



    Azimuth   ■Newly built
              ■Removed
                                 (b) Autumn in 2009

b




                             a

                                                        38
                       Summary
 Remote sensing is quite a promising tool for risk and
  disaster management and its basic issues are highlighted.
 New high-resolution optical sensors can be utilized to
  develop urban inventory and to detect damages caused by
  disasters.
 Thermal sensor and Lidar are useful to gather temperature
  data and to develop DSM/DEM.
 Synthetic Aperture Radar (SAR) can be employed to
  gather spatial data regardless of weather and daylight
  conditions.


                                                            39
Thank you very much!
                       40

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:10/27/2012
language:English
pages:40