aRchi (version 2.1.3)

skeletonize_pc: Build the skeleton of a tree point cloud

Description

Build the skeleton of a tree point cloud

Usage

skeletonize_pc(
  aRchi,
  D = 0.03,
  progressive = TRUE,
  cl_dist = 0.02,
  max_d = 0.05
)
# S4 method for aRchi
skeletonize_pc(
  aRchi,
  D = 0.03,
  progressive = TRUE,
  cl_dist = 0.02,
  max_d = 0.05
)

Value

an aRchi file containing the original point cloud and the corresponding skeleton

Arguments

aRchi

an object of class aRchi containing at least a point cloud.

D

numeric. The distance of research for point neighborhood. Sets the layer thickness. See description for details.

progressive

logical. Should the clustering distance be progressive ? See description for details.

cl_dist

numeric. The clustering distance. If pregressive = FALSE sets the clustering distance for all the point cloud. If pregressive = TRUE sets the minimum clustering distance to be used. See description for details.

max_d

The maximum searching distance for skeleton building. See description for details.

Details

The skeletonization algorithm works in four steps. At STEP 1 the point cloud is divided in layers of regular thickness (defined by parameter D). To do so, the tree base is first defined as the first layer. Then, all the points that are within D of any points of the layer N belong to the layer N+1. This process continues until no more points are found within D. If there are some remaining points (that are further than D to any already classified points), a new layer is defined as the point that is the closest of the already classified point. This continues until no more points remain unclassified. At STEP 2, the layers are divided into clusters based on point distance: two point that are further than a given distance are considered as being part of two different objects. Two possibilities are available to define the clustering distance. First, it remains constant and is defined by cl_dist. Second, the distance is defined as the average distance of the points of the previous layer to the center of their corresponding cluster (this is achieved by setting progressive = TRUE). In this latter case, cl_dist defines the minimal clustering distance that can be used. This option helps to adapt the clustering distance to the size of the actual objects (i.e. branch sections) that have to be to clustered. By default the first layer is assumed as being part of a single cluster. At STEP 3, the cluster centers are used as node of the skeleton and are combined to iteratively build the skeleton. At first iteration, the node with the smallest Z value is used as the root node. At subsequent iterations, the node located within max_d and that is the closest to the root node is integrated into the skeleton and is selected as new root node. This process continues until no node is found within max_d to the root node (either because the cluster is a branch tip or because there is a gap in the point cloud). In this case, the node that belong to the layer that was produced at earliest iteration of STEP 1 and that is the closer to a skeleton node is selected as new root node. This process ends once all nodes are integrated into the skeleton. At STEP 4, a basic tree topology is computed. First, an unique ID is assigned to all segments of the skeleton (i.e. the segment that link two nodes) and its parent segment ID is retrieved. The segments are then classified into axes. An axis being defined as a continuous set of segments that always follow the segment that support the longest portion of the skeleton. Axes are then partitioned into axes segments defined as the portion of an axis located between two branching point or between a branching point and an extremity of the axis. The axis branching order is then computed as the number of branching point observed between the tree base and the axis first segment + 1.

Examples

# \donttest{
#################################################################
# Example with a small high quality point cloud : using default #
# default parameters to detect fine architectural details       #
#################################################################
# import a point cloud
tls=system.file("extdata","Tree_2_point_cloud.las",package = "aRchi")
tls = lidR::readLAS(tls)

# build an empty aRchi file and add the point cloud
aRchi = aRchi::build_aRchi()
aRchi = aRchi::add_pointcloud(aRchi,point_cloud = tls)

# plot the point cloud
plot(aRchi@pointcloud)

# build a skeleton from the point cloud
aRchi = skeletonize_pc(aRchi)

# smooth the skeleton
aRchi = smooth_skeleton(aRchi)

# plot the skeleton
plot(aRchi,show_point_cloud = TRUE)


##############################################################
# Example with a large point cloud with a lot of occlusion : #
# parameters selected for speed                              #
##############################################################
# import a point cloud
tls=system.file("extdata","Tree_1_point_cloud.las",package = "aRchi")
tls = lidR::readLAS(tls)

# build an empty aRchi file and add the point cloud
aRchi = aRchi::build_aRchi()
aRchi = aRchi::add_pointcloud(aRchi,point_cloud = tls)

# plot the point cloud
plot(aRchi@pointcloud)

# build a skeleton from the point cloud
aRchi = skeletonize_pc(aRchi, D = 0.5, cl_dist = 0.2, max_d = 1)

# smooth the skeleton
aRchi = smooth_skeleton(aRchi)

# plot the skeleton
plot(aRchi,show_point_cloud = TRUE)
# }