Talk to be held at the workshop on image analysis and spatial statistics in forestry on November 2, 1999 at KVL, Frederiksberg, Denmark
Ilkka Korpela
(Ilkka.Korpela@helsinki.fi)
Dept. of Forest Resource Management,
University of Helsinki,
Finland
Motivation
Automated interpretation of remote sensing data is tempting in the frame of forest inventory and monitoring. High resolution imagery together with the computing power now at everyone's disposal has provided us with a change in scale with more detailed information being sought through the means of digital image analysis. As an example we can note that highly sophisticated and expensive analog devices are no longer necessarily required in photogrammetry if automated digital photogarmmetry is applied.
Objectives
This project aims at gaining knowledge of the main factors affecting automated tree (~top, ~crown) detection from digital aerial photos. Methods developed by others in similar attempts [e.g. 1,2,3] are being tested to see their validity under varying (Finnish) forest conditions, photo geometry and alternating flying altitude (spatial resolution). These factors have also been studied earlier in the context of visual (non automated) photo interpretation and therefore much can be anticipated from these experiences [e.g. 4, 5, 6, 7]. The project also aims to produce a scientific paper satisfying the requirements of an academic licentiate degree in Forest Mensuration and Management.
Data
Tree-data has been collected from fifteen Scots Pine, Norway Spruce and Silver Birch dominated stands in Hyytiälä in Central Southern Finland 61o51'N 24o20' E . A stand is represented by a 50 x 50 meter plot. Measurements for each tree include location and stem size descriptors. Field measurements have been reported in [8]. Plots represent a silvicultural thinning experiment and include different thinning regimes [9].
Aerial photos were taken in the morning of June 19 1997. They include panchromatic photos in nominal scales 1:10000 and 1:5000 and some color infrared images in 1:16000 and 1:5000. Photos have initially been scanned at 25 um. (25 cm, 12.5 cm) The photos are taken so that every plot (tree species x thinning treatment) would have a 'nadir image', a 'backlighted image' and a 'front lighted image'. Because of restricted funding this criteria could be met only partially. For photo orientation control points have been measured using differential GPS.
Project plan
The project began in 1997 when funding was applied for and received and some empirical data captured. The first two years of the project have been spent in studying subjects underlying digital image analysis and digital photogrammetry.
Phases of the project:
|
i | Measurements for forest data, marking and measuring of control points |
| ii | Aerial photography, acquiring digital images |
| iii | Image orientation, rectification of images & ground truth data, programming a tool for this. (Currently underway) |
| iv | Implementation and testing of "Dralle's method" [2] smoothing technique, Implementation and testing of crown model technique [1,3]. Testing the capability of these algorithms to locate tree tops. Possible 3-D reconstruction of tree crowns from stereo imagery [10,11] |
| v | Reporting the results in the form a licentiate thesis |
1. Kiema P. 1990. Latvusten projisointi numeerisen ilmakuvan koordinaatistoon yksittäisen puun numeerisen ilmakuvatulkinnan menetelmässä. (Projecting images of tree crowns to aerial photo coordinate system in the system of assessing individual trees from numerical photos). Master's thesis. Department of Forest Mensuration and Management. Univ. of Helsinki. In Finnish.
2. Dralle, K. & Rudemo, M. 1996 Stem number estimation by kernel smoothing of aerial photos. Can. J. For. Res. 26: 1228-1236.
3. Larsen M. 1998. Finding an optimal match window for spruce top detection based on an optical tree model. Proceedings of the international forum in Automated Interpretation of High Spatial Resolution Digital imagery for Forestry. Feb 10-12. 1998 Victoria BC. canada. Province of British Columbia. Ministry of Forests. p. 55 - 63.
4. Anttila P. 1998. Analyyttisellä stereoplotterilla ilmakuvilta tulkittujen puukohtaisten tunnusten tarkkuus. (On the accuracy of treewise measurements using analytic stereo plotter). Master's thesis. University of Joensuu. Faculty of Forestry. In Finnish.
5. Ericson, O. 1984. Beståndsinventering med flygbild. Summary: Stand inventory by aerial photogrammetry.
6. Nyyssönen A. 1955. On the estimation of the growing stock from aerial photographs. Metsäntutkimuslaitoksen julkaisuja 46 (1).
7. Nyyssönen, A. , Poso, S. & Keil, C. 1968. The use of aerial photographs in the estimation of some forest characteristics. Acta Forestalia Fennica 82(4).
8. Fish, S., Haakana, M., Korpela, I. Melkas, T. 1998. Spatial dependencies in a forest stand. In: Spatial Statistics in GIS applications. Written reports produced by the students of Postgraduate Course on Geoinformatics. Helsinki University of Technology. Cartography and Geoinformatics 4. p. 7-39.
9. Räsänen P. 1992. Metsänhoitotieteen opetuskoealat Hyytiälässä. Univ. of Helsinki. Dept. of Forest Ecology. Publications 4.
10. Ilvessalo Y. 1950. On the correlation between crown diameter and the stem of trees. MTJ 38 (2).
11. Jacobsons, A. 1970. Sambandet mellan trädkronans diameter
och andra trädfaktorer, främst brösthöjdsdiametern:
analyser grundade på riksskogstaxeringens provträdsmaterial.
Stockholms skoghögsskolan, institutionen för skogstaxering. Rapporter
och uppsatser 14.