www.cfm.ohio-state.eduresearchAIMSION_20GPS.ppt

Document Sample
www.cfm.ohio-state.eduresearchAIMSION_20GPS.ppt Powered By Docstoc
					Information Fusion for Enhanced
      Surface Extraction

 Dorota A. Grejner-Brzezinska and Charles K. Toth
             The Ohio State University
               Center for Mapping

                     ION GPS
               September 14-17, 1999
                Nashville, Tennessee




                           The Ohio State University Center for Mapping
           Presentation Outline

 Data fusion concept
 Multi-sensor systems in mapping
 Airborne laser ranging (ALR)
 Surface extraction with ALR and digital imagery
 DEM extraction and comparison
 Summary, outlook



                                 The Ohio State University Center for Mapping
           When you use information from
           one source, it is plagiarism

           When you use information from
           many, it is information fusion



By: Dr. Belur V. Dasarathy
Editor-in-Chief, Information Fusion
                                      The Ohio State University Center for Mapping
                Information/Data Fusion

Data/information fusion is an engineering process concentrating on
the simultaneous processing of independent data streams
 Goals:
     providing theory, tools and techniques to capitalize on synergy in the
    information from multiple sources
     selecting only a necessary subset of information that would satisfy the task
    at minimum cost
     improving the ability to estimate the characteristic entities and task
    capability
     achieving superior results
 Several levels of fusion


                                                  The Ohio State University Center for Mapping
           Multi-sensor Systems in
                   Mapping
Geometric data fusion  time-space registration
 Basis for higher level fusion (multi-sensor imaging systems)
 GPS/INS integration (complementary and competitive data)
     high accuracy (5-20 cm, 10-30 arcsec)
     high reliability
     fault-resistant
     cost effective (less ground control)
     mandatory for new spatial data sensors (LIDAR, SAR, multi/hyperspectral)
     experimental systems (University of Calgary, OSU AIMS)
     commercial (Applanix)



                                              The Ohio State University Center for Mapping
                                                      GPS Antenna     Imaging PC

                                                                                   INS/GPS PC
                                                           Trimble 4000SSI




 AIMS                 BigShot™ Hasselblad
                             Camera


Prototype
                                                                         LN-100




            GPS Base Station




                                             The Ohio State University Center for Mapping
Problem Definition: Data Fusion for
        Surface Modeling




                      The Ohio State University Center for Mapping
               Problem Statement (1)
 Increased demand for Digital Elevation Model (DEM)
    orthophoto production
    engineering design
    modeling and visualization
    feature extraction

 DEM extraction: feature- and area-based matching
  schemes (image correlation)
    lack of modeling of man-made objects, occlusions, motion artifacts
    good performance for smooth, rolling terrain and small to medium scale
    performance decreases rapidly for more complex scenes and larger scale
    gray-scale stereo image-based surface reconstruction -- virtually
     ill-posed problem

                                               The Ohio State University Center for Mapping
          Problem Statement (2)

 Variety of new spatial data acquisition sensors
    CCD-based cameras
    LIDAR (light detection and ranging)
    multi/hyper-spectral sensors
    SAR/IFSAR
 Optimal sensor fusion
    more consistent scene description
    complementary information
    redundant information


                                         The Ohio State University Center for Mapping
                Airborne Laser Ranging
 Date back to the 1970s and 1980s
 First introduced to the mapping community about a decade ago
 Maturity of GPS/INS direct orientation systems
 Can survey areas and objects difficult to capture or analyze by
traditional photogrammetric methods
     dense city areas,
     power lines,
     forests
     DEM generation for coastal areas and wetlands

 Rapid, highly accurate DEM coverage of medium-sized project areas
 Supports feature extraction and complex 3D modeling

                                             The Ohio State University Center for Mapping
Airborne Laser Ranging




              The Ohio State University Center for Mapping
            Airborne Laser Ranging
 Range accuracy: below 10 to 25 cm
     data form 3D point clusters or lines (3D coordinates)
     elevation has a unique value as a function of the horizontal location
     fast and automated data processing.
 Can deliver multiple echoes from one laser pulse
     separation of terrain or man-made objects from vegetation
 Intensity (reflectance) information
     area classification
 Limitations:
     not capable of any direct pointing to a particular object
     sometimes laser data interpretation can only be performed if the
    oriented image backdrop is available
                                              The Ohio State University Center for Mapping
        Airborne Laser Ranging (ALR)

 Example systems
     USGS system
     AeroScan
     DASA/Dornier
     Optech
     JECA
     SAAB




                               Courtesy of EathData Technologies


                         The Ohio State University Center for Mapping
               Panchromatic Image and Laser
               Elevations Plotted as Intensity




Courtesy of EathData Technologies
                                    The Ohio State University Center for Mapping
                                    LIDAR DEM




Courtesy of EathData Technologies
                                           The Ohio State University Center for Mapping
                Glen Canyon Dam : LIDAR
                         Points




Courtesy of EathData Technologies
                                    The Ohio State University Center for Mapping
         Glen Canyon Dam : LIDAR DEM




Courtesy of EathData Technologies
                                    The Ohio State University Center for Mapping
                Glen Canyon Dam - LIDAR
                                    (10ft contours)




Courtesy of EathData Technologies
                                                      The Ohio State University Center for Mapping
                LIDAR vs Stereo Image
 Assume 1:12,000 scale images and 75 LIDAR FOV (~2000 m altitude)
Typical vertical accuracy
     Stereo-image: 10 to 30 cm
     LIDAR
          ~12 cm at nadir
          ~25 cm at the edges of the scan line (75 FOV)
 The image resolution (and thus the feature measurement accuracy)
decreases with altitude
     ~ 2 m in vertical accuracy at 20,000 m altitude
 LIDAR system still offers (at 20,000 m altitude )
     ~25 cm at nadir
     ~45 cm for the edges of the 75 FOV

                                               The Ohio State University Center for Mapping
   LIDAR DEM vs Stereo Image DEM
 Image-based DEM
     full area coverage
     unable to cope with buildings
     creates smooth, draped surface
     break lines are problematic
 LIDAR DEM
     good point-wise sampling
     surface representation independent from underlying objects
    break lines are problematic due to finite size of a footprint
 Feature extraction needs DEM (possibly from LIDAR)
 Similar pre-processing
     error filtering, data thinning-out
     man-made object removal, and detection of the break lines

                                                 The Ohio State University Center for Mapping
  Test Flight with LIDAR and Digital
         4K by 4K Camera (1)


 Objective: LIDAR and image data collection
 Supported by EarthData Technologies, Hagerstown, MD
 AeroScan LIDAR system
 4K by 4K digital camera
 Differential GPS/INS
 Test area with accurate ground truth




                                    The Ohio State University Center for Mapping
4K by 4K Image of a Test Area
         850 m AGL




                The Ohio State University Center for Mapping
                     Photogrammetric Results

 Aerial triangulation performed for boresight calibration
     1.2 m in horizontal and 0.4 m in vertical directions
     0.02-0.06 degrees for attitude angles
          limited by
               camera resolving power
               asymmetrical target pattern
 OSU-developed hierarchical warped image-based DEM generation
     5 m grid (the closest to the average spacing of the irregular LIDAR data)
     DEM accuracy: 1-1.5 m on average




                                                 The Ohio State University Center for Mapping
               AeroScan LIDAR

 FOV 1-75 deg (300-20,000 ft)
 65 m/s airspeed
 Along track spacing - about 6 m
 Maximum scan rate - 7.5 Hz
 2500 m AGL flying altitude
 Illuminated footprint - 0.6 m
 Typical accuracy on the ground
     0.25-0.35 m cross track
     0.2-0.25 m along track
     0.15-0.25 m height



                                    The Ohio State University Center for Mapping
Distribution of the LIDAR Elevation
 Spots from Different Flight Lines


            Profile 6b




                            Profile 6c




                         The Ohio State University Center for Mapping
   Photogrammetric and LIDAR DEMs
                                    Photogrammetrically-derived DEM
                                    from the 4K by 4K imagery test area




LIDAR DEM observations over
Hagerstown, MD test area




                                         Topographic surface of
                                        Hagerstown, MD test area




Object features of Hagerstown, MD
      test area, NE quadrant.




                                                 The Ohio State University Center for Mapping
                                                                                      Elevation Profiles of LIDAR and
                187

                186
                                                         LIDAR
                                                         Topography
                                                                                      Stereo Image-created Elevation
                         6a
                185
                                                         Photogrammetry
                                                                                          Spots vs Ground Truth
                184
Elevation [m]




                183

                182

                181
                                                                                                        188
                180
                                                                                                        187    6b
                                                                                                                                               LIDAR
                179                                                                                                                            Topography
                                                                                                        186
                                                                                                                                               Photogrammetry
                178                                                                                     185
                  160        170   180      190        200          210   220   230
                                   Distance along the profile [m]
                                                                                                        184




                                                                                        Elevation [m]
                                                                                                        183
                190
                                                                                                        182
                        6c                               LIDAR
                188                                      Topography                                     181
                                                         Photogrammetry
                                                                                                        180
                186
                                                                                                        179
Elevation [m]




                184                                                                                     178
                                                                                                          60        70   80        90        100          110   120   130
                                                                                                                         Distance along the profile [m]
                182


                180


                178
                  60         70    80        90        100          110   120   130
                                   Distance along the profile [m]


                                                                                                                    The Ohio State University Center for Mapping
                         Summary
 Early experiences with LIDAR data for surface
  reconstruction were reported
 Direct georeferencing was used for LIDAR and digital
  camera
 Excellent match of LIDAR DEM with ground truth for
  flat areas
 Significant performance increase can be expected from
  combining LIDAR and image data
    strong geometric constraints provided by LIDAR
    separation of vegetation canopy from topographic surfaces
    processing automation

                                       The Ohio State University Center for Mapping
Courtesy of EathData Technologies
                                    The Ohio State University Center for Mapping
         Areas of Future Research


 DEM comparison
    existing methods do not consider the actual slope at the
     elevation spots
    simultaneous processing of image and LIDAR data
 Automation of LIDAR boresighting
 Use of overlapping LIDAR data
    quality control
    calibration


                                      The Ohio State University Center for Mapping

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:0
posted:3/22/2013
language:English
pages:30