Detection of single trees in aerial images using template matching

Sammanfattning

Today there exist methods for automatic detection of single trees using digitized air photos or remote sensing images with pixel sizes in the order of 0.5 – 1.0 m. By these techniques a majority of trees are detected. In combination with a sample of field plots where all trees are counted, the single tree detection techniques opens up new possibilities in forest inventory and planning. There exist different approaches for the single tree detection, such as segmentation or template matching. The template matching approach was pioneered by Richard James Pollock, University of British Columbia. The idea is that a library of 3-dimensional model trees are cross correlated against any potential tree position in the digital image and tree positions and tree sizes with the highest correlations are considered as likely trees. In this study a computer program based on Pollock’s ideas have been developed and tested on a spruce stand at a test site in Sweden. A technique to remove multiple hits by template coverage is introduced. The program shows promising results and it should be possible to develop an inventory system based on these techniques.

Nyckelord

template matching; single trees; inventory and planning; image processing

Publicerad i

Titel: FORESTSAT 2002: Operational Tools in Forestry Using Remote Sensing Techniques : Conference Papers : August 5th-9th, 2002
Utgivare: Forestry Commission

Konferens

ForestSat 2002, Operational Tools in Forestry Using Remote Sensing Techniques, Heriot Watt University, Edinburgh, August 5th-9th of August 2002

SLU författare

• Olofsson, Kenneth

  • Institutionen för skoglig resurshushållning och geomatik, Sveriges lantbruksuniversitet
    

Associerade SLU-program

Remningstorp

UKÄ forskningsämne

Skogsvetenskap
Jordobservationsteknik

Permanent länk till denna sida (URI)

https://res.slu.se/id/publ/139276