tree_segmentation: Preprocessing and segmentation of raster image for tree...
In lidaRtRee: Forest Analysis with Airborne Laser Scanning (LiDAR) Data

tree_segmentation

R Documentation

Preprocessing and segmentation of raster image for tree identification

Description

global function for preprocessing (filtering), maxima detection and selection, segmentation and segmentation adjustment of a raster image.

Usage

tree_segmentation(dem, dtm = NULL, crown_prop = NULL, crown_hmin = NULL, ...)

Arguments

`dem`	raster object or string indicating location of raster file (typically a canopy height model or a digital surface model; in the latter case the dtm parameter should be provided)
`dtm`	raster object or string indicating location of raster file with the terrain model. If provided, the maxima extraction and watershed segmentation are performed on the dem (this avoids the deformation of crown because of the normalisation with terrain), but maxima selection and segment adjustment are performed on 'dem-dtm' because the selection criteria are based on the height to terrain.
`crown_prop`	(deprecated) numeric. (overrides `prop` parameter passed to `seg_adjust`, for backward compatibility)
`crown_hmin`	(deprecated) numeric. (overrides `min.value` parameter passed to `seg_adjust`, for backward compatibility)
`...`	arguments passed to functions `dem_filtering` (e.g. `nl_filter`, `nl_size`, `sigma`), `maxima_detection`, `maxima_selection`, `maxima_selection` (`dmin`: treetop minimum distance to next higher pixel in meters, `dprop`: number defining the treetop minimum distance as proportion of its height to next higher pixel, `hmin`: minimum treetop height), `seg_adjust` (`prop`: minimum height of tree crown base as proportion of treetop height, `min.value`: minimum crown base height)

Value

A SpatRaster with 4 layers: selected local maxima (values = distance to higher pixel), segments, non-linear preprocessed dem, smoothed preprocessed dem

References

Monnet, J.-M. 2011. Using airborne laser scanning for mountain forests mapping: Support vector regression for stand parameters estimation and unsupervised training for treetop detection. Ph.D. thesis. University of Grenoble, France. Section 6.2 https://theses.hal.science/tel-00652698/document

Monnet, J.-M., Mermin, E., Chanussot, J., Berger, F. 2010. Tree top detection using local maxima filtering: a parameter sensitivity analysis. Silvilaser 2010, the 10th International Conference on LiDAR Applications for Assessing Forest Ecosystems, September 14-17, Freiburg, Germany, 9 p. https://hal.science/hal-00523245/document

Examples

data(chm_chablais3)
chm_chablais3 <- terra::rast(chm_chablais3)

# tree segmentation
segments <- tree_segmentation(chm_chablais3)
segments2 <- tree_segmentation(chm_chablais3,
  nl_filter = "Median", nl_size = 3,
  sigma = cbind(c(0.2, 0.8), c(0, 15)), dmin = 0, dprop = 0, hmin = 10, 
  crown_prop = 0.5, crown_hmin = 5
)

# plot initial image segments
terra::plot(chm_chablais3, main = "Initial image")
terra::plot(segments$smoothed_dem, main = "Filtered image")
terra::plot(segments$local_maxima, main = "Local maxima")
#
# replace segment with id 0 (not a tree) with NA
segments$segments_id[segments$segments_id == 0] <- NA
terra::plot(segments$segments_id %% 8, main = "Segments", col = rainbow(8))
#
# plot segmentation with other parameters
segments2$segments_id[segments2$segments_id == 0] <- NA
terra::plot(segments2$segments_id %% 8, main = "Segments2", col = rainbow(8))

lidaRtRee documentation built on April 3, 2025, 10:29 p.m.