itd_lmf function

Individual Tree Detection Algorithm

This function is made to be used in locate_trees . It implements an algorithm for tree detection based on a local maximum filter. The windows size can be fixed or variable and its shape can be square or circular. The internal algorithm works either with a raster or a point cloud. It is deeply inspired by Popescu & Wynne (2004) (see references).

lmf(ws, hmin = 2, shape = c("circular", "square"), ws_args = "Z")

Arguments

• ws: numeric or function. Length or diameter of the moving window used to detect the local maxima in the units of the input data (usually meters). If it is numeric a fixed window size is used. If it is a function, the function determines the size of the window at any given location on the canopy. By default function takes the height of a given pixel or point as its only argument and return the desired size of the search window when centered on that pixel/point. This can be controled with the ws_args parameter
• hmin: numeric. Minimum height of a tree. Threshold below which a pixel or a point cannot be a local maxima. Default is 2.
• shape: character. Shape of the moving window used to find the local maxima. Can be "square" or "circular".
• ws_args: list. Named list of argument for the function ws if ws is a function.

Examples

LASfile <- system.file("extdata", "MixedConifer.laz", package="lidR")
las <- readLAS(LASfile, select = "xyzi", filter = "-inside 481250 3812980 481300 3813050")

# =================
# point-cloud-based
# =================

# 5x5 m fixed window size
ttops <- locate_trees(las, lmf(5))

#plot(las) |> add_treetops3d(ttops)

# variable windows size
f <- function(x) { x * 0.07 + 3}
ttops <- locate_trees(las, lmf(f))

#plot(las) |> add_treetops3d(ttops)

# Very custom variable windows size
f <- function(x, y, z) { x * 0.07 + y * 0.01 + z}
ws_args <- list(x = "Z", y = "Intensity", z = 3)
ttops <- locate_trees(las, lmf(f, ws_args = ws_args))

# ============
# raster-based
# ============

chm <- rasterize_canopy(las, res = 1, p2r(0.15), pkg = "terra")
ttops <- locate_trees(chm, lmf(5))

plot(chm, col = height.colors(30))
plot(sf::st_geometry(ttops), add = TRUE, col = "black", cex = 0.5, pch = 3)

# variable window size
f <- function(x) { x * 0.07 + 3 }
ttops <- locate_trees(chm, lmf(f))

plot(chm, col = height.colors(30))
plot(sf::st_geometry(ttops), add = TRUE, col = "black", cex = 0.5, pch = 3)

References

Popescu, Sorin & Wynne, Randolph. (2004). Seeing the Trees in the Forest: Using Lidar and Multispectral Data Fusion with Local Filtering and Variable Window Size for Estimating Tree Height. Photogrammetric Engineering and Remote Sensing. 70. 589-604. 10.14358/PERS.70.5.589.

See Also

Other individual tree detection algorithms: itd_manual