treecbh

==========

The treecbh package provides functions to detect individual tree level Crown Base Height (CBH) using high-resolution LiDAR data. Individual tree segmentation must be conducted prior. The package is meant to be used within the framework of the lidR package.

==========

Required packages

Make sure that the following R packages are installed before using treecbh:

tidyverse, lidR, sf, data.table, crayon, factoextra, fpc, geometry, gtools.

Installation

devtools::install_github("DijoG/treecbh")

Example

Demonstration of how to use treecbh combined with lidR.

Data preparation

require(lidR);require(tidyverse)

# Forest point cloud (low resolution example data)
LAS <- system.file("extdata", "MixedConifer.laz", package = "lidR")
Alas <- readLAS(LAS, filter = "-drop_z_below 0") 

# Black, white and green color palette for visualizing CHM
bgcol <- function(x)
{
  col <- grDevices::colorRampPalette(c("grey1", "white", "forestgreen"))
  return(col(x))
}

Computing canopy height model

Using the pitfree algorithm.

CHM <- rasterize_canopy(Alas, 0.5, pitfree(subcircle = 0.25))
plot(CHM, main = "CHM 0.5 pitfree", col = bgcol(50))

Computing treetops

Using a constant windows size of 5 m.

ws <- 5
treetops <- locate_trees(CHM, lmf(ws))
plot(CHM, main = "CHM 0.5 pitfree", col = bgcol(50))
plot(sf::st_geometry(treetops), add = T, pch = "+", col = "firebrick3")

Point-based crown segmentation

Using the Dalponte algorithm.

algo_crowns <- dalponte2016(CHM, treetops)
las_crowns <- segment_trees(Alas, algo_crowns, attribute = "ID")

las_crowns is a las object storing the ids of individual trees (ID attribute). This object can also be fed into the treecbh::get_CBH() function after storing each individual point cloud (tree) in a directory. Since there are other point- and CHM-based its algorithms outside the liDR framework, trecbh::get_CBH() accepts las files that can previously be isolated based both on points and CHMs.

Obtaining individual tree segments (its) as polygons.

Apoly <- crown_metrics(las_crowns, attribute = "ID", geom = "concave", func = NULL)
plot(sf::st_geometry(Apoly), reset = FALSE, col = "forestgreen", border = "grey80")
plot(sf::st_geometry(treetops), add = T, pch = "+", col = "firebrick3")

Some trees were not segmented.

Extracting point clouds to its (classifying ground points and saving las files by default)

This step is only necessary if you have different its shape file coming from an outer source.

oudir <- "<path to directory>"
treecbh::get_3DTREE(Alas, Apoly, output_dir = oudir, FEATURE= "ID", RETURN = F)

CBH detection

Optimization deactivated, performing treeiso plus cbh detection

Parameter adjustment: cbh_ONLY = 1. Let's proceed with five point clouds of deciduous trees to be found in the https://github.com/DijoG/storage/tree/main/cc_lasfin directory.

?treecbh::get_CBH()

# Example: list of high resolution individual las tree clouds, stored normally in 'outdir' set in the previous step.
# example files used, are in https://github.com/DijoG/storage/tree/main/cc_lasfin library
las_l <- list.files(oudir, pattern = ".las", full.names = T) %>%
  gtools::mixedsort()

# visualizing one
plot(readLAS(las_l[2]), bg = "white", size = 5, axis = T)

# output directories
outdi1 <- "<path to directory>" # treeiso isolated tree cloud
outdi2 <- "<path to directory>" # filtered tree cloud (stem plus first leaved branch)

# CC executable 
cc_dir <- "<path to /CloudCompare.exe>"

# running 
A_CBH <- treecbh::get_CBH(las_l,
                          # run tree isolation and cbh detection
                          cbh_ONLY = 1,
                          # its point cloud directory
                          outdir1 = outdi1,
                          # stem plus first leaved branch directory
                          outdir2 = outdi2,
                          cc_dir = cc_dir)
A_CBH

Output parameters:
Z_max → Maximum height of within-segment point cloud (unit: m),
Z_mean → Mean height of within-segment point cloud (unit: m),
Z_sd → Standard deviation of within-segment point cloud (unit: m),
Z_N_points → Height of maximum binned count (unit: m),
N_points → Number of points in the maximum binned count (unit: n),
CBH → Detected Canopy Base Height (unit: m) → Detected Canopy point cloud,
Hull_area → Area of Delaunay convex hull of Detected Canopy point cloud (unit: m²),
Del_vol → Volume of Delaunay convex hull of Detected Canopy point cloud (unit: m³),
Cube_vol → Volume of voxelized (0.2 m) Detected Canopy point cloud (unit: m³),
Sphere_vol → Volume of in-voxel-fit spheres (diameter 0.2 m) of Detected Canopy point cloud (unit: m³),
treeID → initial individual tree cloud id.

Optimization deactivated, executing treeiso only

Parameter adjustment: cbh_ONLY = 2.

treecbh::get_CBH(its_l,
                 outdir1 = outdi1,
                 outdir2 = outdi2,
                 # run only tree isolation (disabling cbh detection)
                 cbh_ONLY = 2,
                 cc_dir = cc_dir)

Optimization activated, executing cbh detection only

Parameter adjustment: cbh_ONLY = 3 and kM = TRUE (k-Means Clustering set to TRUE). The interactive visualised optimization process starts with: 'Do you accept k?' and a plot of recommended k clusters. User answers, if the answer is no, the user is asked to type a number for k after 'Enter k:'. It can be repeated twice.

O_CBH <- treecbh::get_CBH(its_l,
                          outdir1 = outdi1,
                          outdir2 = outdi2,
                          # run only cbh detection (disabling treeiso)
                          cbh_ONLY = 3,
                          # Optimization activator
                          # kM = FALSE is default
                          kM = TRUE,
                          # inactive CC executeable directory
                          cc_dir = cc_dir)

Example 02.las (see R Console right) plots are displayed:

O_CBH