Revision as of 23:48, 6 January 2018

Filter the CHM derived from LiDAR data

We use a Canopy Height Model (CHM) derived from LiDAR data as decribed here to detect Individual Tree Crowns (ITC). Two preprocessing steps prepare a watershed segmentation approach: (1) Gaussian filtering and (2) inversion of a CHM.

• In the search engine of the Processing Toolbox, type smooth and select Smoothing (gaussian) under Image filtering of the Orfeo Toolbox.
• Select the CHM raster data file in GeoTiff format as input layer.
• The smoothing type is gaussian.
• The circular structuring element has a radius of 2 pixels.
• Enter name and path for an output file.
• Click on Run.

Invert the CHM

Now the smoothed CHM will be inverted.

• In the search engine of the Processing Toolbox, type invert and select Invert grid under Raster tools of SAGA.
• Select the smoothed CHM raster data file from previous step as input layer.
• Enter name and path for an output file.
• Click on Run.

Watershed segmentation

• In the search engine of the Processing Toolbox, type watershed and select Watershed segementation under Image Analysis of SAGA.
• Select the inverted and smoothed CHM raster data file as input Grid.
• The Output is Segment ID
• Select as Method the flow accumulation of Minima
• Seed points: enter name and path for a vector point output file.
• Click on Run.

Extracting tree heights

We extract normalized height from the original CHM using the QGIS point sampling plugin.

Generate a seed grid

Seeded region growing