R – Using lidR to Classify Powerline Features Misclassified as Noise

classificationlidarlidrr

I am using NOAA LiDAR data to experiment with methods in lidR. I want to use the linear segment_shapes() function to locate powerlines, but some of powerlines are already classified as noise. Using lidR (preferably) or unlicensed LAStools &/or QGIS how do I reset the classification of these points.

I have used the classify_noise() function in both the las and catalog format. I can't find a simply way to view classifications.

Otherwise is it possible to tell the function from ch 13 in the lidR textbook to consider all points despite already being classified?

lasfile <- readLAS(here("ch13/central_blvd_palmbeach/91_10/9110/Job648155_26080_91_10.las"))
las_denoise <- classify_noise(lasfile, ivf(5,2))
seg_las <- segment_shapes(las_denoise, shp_line(th = 8, k = 15), "linear")
plot(seg_las, color = "Classification")
writeLAS(seg_las, here("ch13/central_blvd_palmbeach/91_10/9110/seg_las.las"))

I am also having trouble updating the classification based on segment_shapes() algorithm results.

screen shot of LiDAR data from NOAA with powerlines. Yellow represents noise classification which is how the dataset came from NOAA

Best Answer

How do I reset the classification of these points.

You can change the classification manually like in most R object

las$Classification[las$Classification == LASNOISE] <- LASUNCLASSIFIED
# if you prefer a data.table style
las@data[Classification == LASNOISE, Classification := LASUNCLASSIFIED]

I can't find a simply way to view classifications.

You can use plot(las, color = "Classification"), noise point are yellow or you can e.g. filter the noise points filter_poi(las, Classification == LASNOISE).

I am also having trouble updating the classification based on segment_shapes() algorithm results

las <- segment_shapes(las, shp_line(th = 8, k = 15), "linear")
las@data[linear == TRUE, Classification := LASWIRECONDUCTOR]

Otherwise is it possible to tell the function from ch 13 in the lidR textbook to consider all points despite already being classified

It already considers all points, But you can use the filter argument to do not consider all points

I want to use the linear segment_shapes() function to locate powerlines

Looking at your point cloud I hardly believe that you will succeed in segmenting powerline that way. The powerlines are sparse and discontinuous due to low sampling density and thus an eigen decomposition is unlikely to detect linear feature in isolated points.

Related Solutions

lidR – Resolving Issues with Classifying by Shapefile Using merge_spatial in lidR

You were almost there but you made some R errors. Below a full reproducible example

shp_merge_spatial = function(las, shp) UseMethod("shp_merge_spatial", las)

shp_merge_spatial.LAScluster = function(las, shp) {
  # The function is automatically fed with LAScluster objects
  # Here the input 'las' will a LAScluster
  
  las <- readLAS(las)                          # Read the LAScluster
  if (is.empty(las)) return(NULL)              # Exit early (see documentation)
  
  las <- merge_spatial(las, shp, attribute = "in_poly")      # merge_spatial
  las$Classification[las$in_poly] <- LASBUILDING           # classify
  return(las)                             # Return the point cloud
}

shp_merge_spatial.LAScatalog = function(las, shp){
  opt_chunk_buffer(las) <- 0
  catalog_sapply(las, shp_merge_spatial, shp = shp)
}

library(lidR)

LASfile <- system.file("extdata", "Megaplot.laz", package="lidR")
ctg = readLAScatalog(LASfile, chunk_size = 150)

shp   <- system.file("extdata", "lake_polygons_UTM17.shp", package = "lidR")
lakes <- sf::st_read(shp)

opt_output_files(ctg) <-"{tempdir()}/{ID}_bldg_cls"
output <- shp_merge_spatial(ctg, lakes)

plot(output)
las = readLAS(output)
plot(las, color = "Classification")

This is a bit complex because merge_spatial does not have LAScatalog variant. But in lidR 3.2.0 that will be released probably next week there is classify_poi()

library(lidR)

LASfile <- system.file("extdata", "Megaplot.laz", package="lidR")
ctg = readLAScatalog(LASfile, chunk_size = 150)

shp   <- system.file("extdata", "lake_polygons_UTM17.shp", package = "lidR")
lakes <- sf::st_read(shp)

opt_chunk_size(ctg) <- 150
opt_output_files(ctg) <-"{tempdir()}/{ID}_bldg_cls"

output = classify_poi(ctg, LASBUILDING, roi = lakes)

plot(output)
las = readLAS(output)
plot(las, color = "Classification")

lidR Package Output – Where Results are Stored When opt_output_files() is set to Empty

When the successive outputs are not written to disk they are stored in R memory as regular R objects like any classical code.

When a function processing a catalog fails midway it does not actually fails programmatically speaking and it returns a partial output. In your case metrics_w2w is not empty which would have been the case in an actual failure scenario.

With opt_restart(bigCat) = 42 you can redo the computation starting at chunk 42. However there is no function to merge the first partial output with the second. You must do some processing yourself. But at least the 42 first chunks are not supposed to be lost.

Best Answer

Related Solutions

lidR – Resolving Issues with Classifying by Shapefile Using merge_spatial in lidR

lidR Package Output – Where Results are Stored When opt_output_files() is set to Empty

Related Question