How to filter heights using LAScatalog in lidR

Question

I would like to filter the points that are used in a canopy height model. In particular I would like filter out any points whose height is over 200 feet. I've tried the code below but when I plot the difference between my normalized and filtered tiles, they are exactly the same and there are still z values above 200 feet. I must be using the opt_filter command incorrectly, or have chosen the wrong mechanism.

library(lidR)

indata <- suppressWarnings(readLAScatalog ("c:/test/laz"))

# Normalize point cloud data
opt_output_files(indata) <- "c:/test/norm/norm_{XRIGHT}_{YBOTTOM}"
normalized_las <- normalize_height (indata, tin(), na.rm=TRUE)

# Keep points where height is between 0 and 200 feet
filtered_las <- suppressWarnings(readLAScatalog ("c:/test/norm/"))
opt_filter (filtered_las) <- "Z>=0 and Z<=200"
opt_output_files(filtered_las) <- "c:/test/filt/filt_{XRIGHT}_{YBOTTOM}"
filt_ctg <- catalog_retile(filtered_las)

# Now read in the same filtered and normalized tile
filt1 <- 'c:/test/filt/filt_2885000_2320000.las'
filtlas <- readLAS(filt1)
norm1 <- 'c:/test/norm/norm_2885000_2320000.las'
normlas <- readLAS(norm1)

# Visualize difference between filtered and unfiltered points
par(mfrow=c(2,2))
boxplot(normlas@data$Z, ylim = c(-10,1000), ylab = "Height (ft)", main = "Before Filtering")
boxplot(filtlas@data$Z, ylim = c(-10,1000), ylab = "Height (ft)", main = "After Filtering")
hist(normlas@data$Z, xlim=c(-2,1000), xlab = "Height (ft)", main = "Before Filtering")
hist(filtlas@data$Z, xlim=c(-2,1000), xlab = "Height (ft)", main = "After Filtering")

Answer 1

You are mistaking the argument filter in readLAS() and the function filter_poi().

filter_poi()

When you have read a LAS file in R and the point cloud is already loaded, you can use a classical R syntax, that you are likely to be familiar with because dplyr popularize it a lot, to select some points of interest (POI)

las = readLAS(LASfile)
las = filter_poi(las, Z>=0, Z<=200)

filter argument in readLAS()

readLAS can take an optional parameter filter and allows selection of “rows” (or points) while reading the LAS file. It happens at C++ level and not at the R level. It is better for different reasons that are documented in readLAS() and in this vignette. But because it does not happen at R level, R syntax does not work. Instead you have to use predefined set of commands. List of commands can be printed using readLAS(filter = "-help"). You can do:

las = readLAS(LASfile, filter = "-drop_z_below 0 -drop_z_above 200")

opt_filter()

When processing a LAScatalog you do not call readLAS() yourself. It is called internally.opt_filter() is used to propagate the information:

opt_filter(ctg) = "-drop_z_below 0 -drop_z_above 200"

Other filter

For the sake of comprehensiveness I must also speak about the other filter argument you can find in lidR such as in grid_metrics() or segment_shapes(). Those filter arguments are R-based filters and thus expect R syntax. They allow for processing a subset of the point-cloud without creating a copy of the point-cloud unlike with filter_poi(). They correspond to clever memory optimization but cannot exist for every function.

metrics = grid_metrics(las, ~mean(Z), 20, filter = ~Z>=0 & Z<=200)

Summary

1. Use readLAS(... filter = "-command") whenever you can (i.e. when it exists a command that do what you want to do) to do not load in R the points you don't need.
2. Use filter = R expression if you want to use filter_poi() but the function you are using is capable of being smarter than that.
3. Use filter_poi() the other cases.
4. Read the LAS formal class vignette and/or the first chapter of the lidR book.