Future Goals

Data Base

I want to setup a database to store point clouds with additional meta data. The metadata should at least contain:

  • Tree Species
  • Tree age
  • Ground Truth data if available
  • Publication
  • Contact
  • Spatio temporal stamps
The SimpleTree program should provide an interface to access the data base directly as well as the homepage.

Ground Modelling and automatic tree cloud extraction

This should be done easily in some cases and others cases will require more work. Most of the scans I use are taken from an agro forestry system. The terrain is flat and contains no shrubs or other objects. Trees do not intersect so much, occlusion is low.

I also have scan data taken in steeper terrain with lots of shrubs and boulders ect. Here a fully automated ground detection might be more difficult. Intersecting tree crowns will make it harder to extract tree clouds separately. One approach might be to just detect the ground points and remove them. If memory allows this, the whole cloud could be processed at once after this. If you are familiar with my method, it should be sufficient to initialize my algorithm with multiple stem bases instead of only one. Then in a first loop all the stems would be detected, in a second loop all first order branches, then all second order branches. This approach should prevent the detection of endless branches jumping over multiple crowns.

Tree species detection

Point feature histograms provided by PCL might be a promising approach to detect tree species efficiently. One could compute those on DBH slices of trees and further process the output with traditional remote sensing classification methods.

More Automatism

Thresholds used in this software require knowledge of how my method works. The algorithms need to be understood to perform the thresholding task efficiently. It would be desirable if the program includes more automatism to make it easier to produce fast reliable results with it. Tree species detection and ground modelling are here a dependency in my opinion.

R package

It would be nice to develop a well formed interface with R. Statistical methods on my tree models should be wrapped inside a R- package.

Time series analysis

Research on time series of TLS derived tree models is envisaged.

 
This template downloaded form free website templates