Skip to main content
SLU publication database (SLUpub)

Report2003Open access

TreeD version 0.8 : an image processing application for single tree detection

Olofsson, Kenneth

Abstract

The TreeD © software, version 0.8, is an application that loads an aerial or satellite image and detects trees by template matching. The software is developed at the Remote Sensing Laboratory, Swedish University of Agricultural Sciences ( SLU) in Umea, Sweden. This is beta software and the methods and algorithms are continuously being improved upon. Please report any errors you find in the software to the Remote Sensing Laboratory. The algorithms for tree detection used in the application are based on a PhD thesis by Richard J ames Pollock (1996), University of British Columbia, Canada. The application is build upon two software libraries, Intel® Image Processing Library, IPL ©and wxWindows ©. The I PL is used for image processing and wx Windows is used as a graphical user interface. The report contains three major sections, a manual, a method description and a software description. The manual is for someone, without any prior knowledge of template matching, who wants to run the software. The method and software descriptions are for someone that wants to build a similar application as Tree D or as a support for in-house development at the Remote Sensing Laboratory. The TreeD 0.8 application is available at the Remote Sensing Laboratory, SLU and is primarily used as a research tool for single tree detection. The software runs on a Microsoft® Windows 2 000 workstation. The application binary depends on the Intel® Image Processing Library, IPL © DLLs and consequently they need to be put into the same folder as the program. All of these binaries can be found at SLU. The input to the application is an aerial/satellite image (central or orthogonal projection), a tree library, information about the camera and solar positions , and a path to a directory to put the results in. The current status of the input variables can be viewed and changed before starting the correlation of the image. The output from the application consists of three text files , status. txt, treelist.txt and probable_treelist.txt. I f there are old files with these names at the result directory they will be overwritten. To save a new batch you can either rename the oldfiles or use a new result directory. ·· The application assumes that the terrain is fairly flat and that the camera is positioned approximately in a nadir view. I f these conditions are not fulfilled the accuracy of the positioning and detection of the trees will decrease.

Keywords

trees; computer software; aerial surveying

Published in

Arbetsrapport / Sveriges lantbruksuniversitet, Institutionen för skoglig resurshushållning
2003,
Publisher: Institutionen för skoglig resurshushållning, Sveriges lantbruksuniversitet