Ilkka Korpela - Work / Research
This page appeared first in Dec 28, 2005 - Last modified Nov 13, 2012.
 Pages: [Index, Research, Teaching, Recent   |   Family, Botanics, Amateur radio, Links, Webmail]

   
Logo (3D Treetop positioning)     Year 2004 at geodetic point 88M5041 during LiDAR campaign      Lapinkangas 2011 - canopy photography control points

ONGOING AND RECENT

CV and List of Publications.  Presentations on research you can find below and
teaching material from the teaching page. 
The WebLog provides with daily-weekly updates.
My current interest is 3D remote sensing using optical data

Keywords: Forestry, Forest inventory, Forest management, Applied botanics, Allometry, Mapping, Photogrammetry, LiDAR, Analytical geometry, Radiometry,  Programming, Numerical methods, Monte-Carlo techniques, Methods of least squares, Sampling, Linear models.
DATA-DRIVEN RESEARCH CALLS FOR OBSERVATIONS -
EXPERIMENT IN HYYTIÄLÄ

I've maintained an experiment in Hyytiälä (62ºN, 24ºE) since 1993.
It consists of remote sensing and field observations.

Description and links to data [HTML in Finnish].[pdf in English].

1. There are accurately mapped trees in different forests; Characteristic to the  experiments is that all observations are georeferenced.
2. Aerial image time-series covers all photographs since 1946, scale > 1:32000.
Last film images are from 2004. Digital images prevail since then.

1946 vasen 1946 oikea

1946 pair, normalized, as Schenk defines it.
3. Airborne laser scanning: There are now eight LiDAR data sets: 2004, 2006, 2007, 2008, 2010, 2011 (2 sets) and 2012.
 
4. A set of 'geodetic points' has been established in the area since 2000 using ordinary RTK and  Network RTK with an accuracy of 2-3 cm.

5. Field photogrammetry: Close-range photogrammetric images since 2007. 

The forests are managed by Metsähallitus.
GOOGLE-HYYTIÄLÄ

Hyytiälä is covered by a time-series of aerial images starting in 1946, which are all in one large block that has been determined accurate orientation. Here's a pdf by Silva Fennica describing how it has been done. With the imagery is possible to do ´time-travelling in 3D`. It is exciting to combine lidar with photogrammetry. See video-clips of the laser scanning .(3.6 Mb Ground points) (5.5 Mb  Points >20 m fm ground).

In the summer of 2006, we set up the first version of the 'GOOGLE-HYYTIÄLÄ' (cf. GoogleEarth, :)) web-site and interviewed people in Hyytiälä about the landscape changes. Aerial photographs have wittnessed a big change in the land use patterns in Hyytiälä. Please visit the www site, the images are not in Finnish.

KUVAMITT ENDEAVOURS

KUVAMITT's display of surface models

In 1999, I wrote the first version of a VB 6.0 program that could perform multi-image matching, i.e. least-square forward ray-intersection for 3D point positioning. At the time I had realized that I almost completely lack stereo vision.

KUVAMITT is written in VStudio 6.0 (Win32) that provides the user-interface. Computer intensive calculations are done in a C/C++ dll. It is linked with the WIN32 API and Matlab dlls. VB 6.0 by Microsoft (1998-) is still somewhat alive and the support is promised to last until 2018. KUVAMITT's principle is that any step in an analysis that can be visualized is visualized. This is especially in the favor of using it for teaching photogrammetry.

KUVAMITT's features include
  • Image triangulation in a weighted least-square (bundle block) adjustment, All observations have weights, reflecting the a priori observation accuracy. image-, XYZ-, XY- or Z-constrained points, and sensor position/attitudeobservations. 
  • Monoscopic, manual multi-image matching of upto 9 images.
  • Stereocopic mode in anaglyph. View and simple editing.
  • Ortoview mode, Fast switching between mono, stereo, orto views.
  • Digitizing of vectors in all modes, no topology so far, just points and polylines.
  • Fly-through videos: given path output an image sequence.   
  • Use of two surface models, DTM/CHM, ebinary raster or ASCII TIN.
  • Orthoimage computation with a raster surface model.
  • Semi-automatic image matching for 3D tree top positions by template matching
  • Image-based crown-width estimation using multi-scale template matching
  • Tree crown 3D modeling using parametric crown models and LiDAR points with allometric constraining
  • Area-based stereo- and multi-image surface matching into a TIN. Routines for filtering the TIN for outliers and outputting a raster surface model.
  • Superimposing 3D vectors in all view-modes: mono, stereo, orto
  • I/O of 1-4 band images, 8-bit and 16-bit. RAW (BIP) images only. ASCII headers.
  • Film image, Vexcel UltracamD/X, DMC I/II, and ADS40 sensor support.
  • Scene shadow and occlusion determination using Z buffering in LiDAR data
If you would be interested in developing KUVAMITT into an "academic freeware" DPW, please contact me by e-mail. There is a 5 gB sample package and instructions in docx 



SPONSORS (CURRENT AND PAST SPONSORS OF  EXPERIMENTATION)
 
       
  '  
    


 
     
 


Older presentations (not updated)