10 Point Cloud Analysis

Published

January 1, 2026

The Reality Capture API can output a point cloud alongside (or instead of) a mesh. Once you’ve converted it to LAS format, lidR (Roussel et al. 2020) takes over, height statistics, density maps, canopy models, voxel volumes. Visualisations use ggplot2 (Wickham 2016). This is the go-to tool for site survey analysis.

10.1 Getting the Point Cloud

Request rcs format when creating your photoscene:

library(AutoDeskR)

resp    <- getToken(id     = Sys.getenv("client_id"),
                    secret = Sys.getenv("client_secret"),
                    scope  = "data:read data:write")
myToken <- resp$content$access_token

ps <- createPhotoscene(name = "site-2026-q2", token = myToken)
# ... uploadImages(), processPhotoscene(), waitForPhotoscene() ...

result  <- checkPhotoscene(photoscene_id = myPhotosceneId, token = myToken)
rcs_url <- result$content$photoscene$scenelink

download.file(url      = rcs_url,
              destfile = "site-survey.rcs",
              headers  = c(Authorization = paste("Bearer", myToken)))

Warning

lidR reads LAS/LAZ, not RCS. Convert the downloaded .rcs file to LAS using CloudCompare (free and open source) before proceeding. In CloudCompare: File → Open the RCS file, then File → Save As → LAS format.

10.2 Reading the Point Cloud

library(lidR)
pc <- readLAS("site-survey.las")
pc
#> class        : LAS (v1.3 format 1)
#> memory       : 247.8 Mb
#> extent       : 406325.4, 406412.8, 5765842.6, 5765901.3
#> coord. ref.  : WGS 84 / UTM zone 32N
#> area         : 4064.9 units²
#> points       : 3.62 million pts (1 return)

10.3 Height Statistics

Normalise to ground level before computing anything height-related:

pc_norm <- normalize_height(pc, knnidw())

mean(pc_norm$Z)
#> [1] 4.12    # mean height above ground (m)
sd(pc_norm$Z)
#> [1] 2.87

quantile(pc_norm$Z, probs = c(0.25, 0.5, 0.75, 0.95))
#>   25%   50%   75%   95%
#>  1.20  3.44  6.11 11.83

10.4 Height Distribution

Figure 10.1 shows the distribution of normalised point heights. The bulk of returns between 0–2 m are ground-level; the secondary peak around 8–12 m corresponds to roof surfaces.

ggplot(data.frame(z = z_sample), aes(x = z)) +
  geom_histogram(binwidth = 0.25, fill = "#367ABF", colour = "white") +
  labs(title = "Point Cloud Height Distribution",
       subtitle = "Normalised above ground level",
       x = "Height (m)", y = "Point count") +
  theme_minimal()

Figure 10.1: Height distribution of normalised point cloud returns (above ground level).

10.5 Canopy Height Model

Rasterise the maximum Z per cell to produce a canopy height model, great for visualising roof heights or vegetation structure across the site:

chm <- rasterize_canopy(pc_norm, res = 0.5, algorithm = p2r())
plot(chm, col = height.colors(50), main = "Canopy Height Model (0.5 m resolution)")

10.6 Point Density Grid

density <- grid_density(pc, res = 1)
plot(density, col = gray.colors(50), main = "Point Density (pts/m²)")

mean(density[], na.rm = TRUE)
#> [1] 891.3   # points per square metre

10.7 Voxel Volume

vox        <- voxelize_points(pc_norm, res = 0.25)
volume_m3  <- nrow(vox) * 0.25^3
cat("Estimated volume:", round(volume_m3, 1), "m³\n")
#> Estimated volume: 1483.2 m³

10.8 Summary Table

library(ggplot2)

metrics <- data.frame(
  metric = c("Total points", "Survey area (m²)", "Mean density (pts/m²)",
             "Mean height (m)", "95th-pct height (m)", "Voxel volume (m³)"),
  value  = c(npoints(pc), area(pc), mean(density[], na.rm = TRUE),
             mean(pc_norm$Z), quantile(pc_norm$Z, 0.95), volume_m3)
)

ggplot(metrics, aes(x = reorder(metric, value), y = value)) +
  geom_col(fill = "#367ABF") +
  coord_flip() +
  labs(title = "Point Cloud Summary Metrics",
       x = NULL, y = "Value") +
  theme_minimal()