segmentation: Segmentation

In ggRaver/Lslide: Functions for object-oriented image analysis in landslide science

Segmentation

Description

This function uses the open-source software of GRASS and SAGA GIS to fulfill an imagery segmentation. It is possible to decide bet ween SAGA and GRASS GIS segmentation.

In SAGA GIS, the tools SEED GENERATION or FAST REPRESENTATIVENESS are used for computing seed points. Then, the SEEDED REGION GROWING al- gorithm is used for the segmentation.

In GRASS GIS, the tool OBJECT-SEGMENTATION (i.segment) is used for the segmentation. There is also the possibility to use the new SLIC algorithm.

Moreover, there is the option to compute a generalisation of segments by (multiple) majority filter (SAGA GIS) at the end.

Usage

segmentation(Tool, Segments.Grid, Segments.Poly, Input.Grid,
  Saga.Output.Grid = file.path(tempdir(),
  paste0("SagaRepresentativenessOutputGrid", par.i, ".sgrd")),
  Saga.Output.Lod = file.path(tempdir(),
  paste0("SagaRepresentativenessOutputLod", par.i, ".sgrd")),
  Output.Seeds = file.path(tempdir(), paste0("OutputSeed", par.i,
  ".sgrd")), Fast.Representativeness.LevelOfGeneralisation = "10.0",
  Saga.Similarity = file.path(tempdir(), paste0("SagaSimilarity", par.i,
  ".sgrd")), Saga.Segments.Seeds.Table = file.path(tempdir(),
  paste0("SagaSegmentsSeedsTable", par.i, ".mtab")),
  Saga.Segmentation.Normalize = "0",
  Saga.Segmentation.Neighbourhood = "0",
  Saga.Segmentation.Method = "0", Saga.Segmentation.Sig.1 = "1.0",
  Saga.Segmentation.Sig.2 = "1.0", Saga.Segmentation.Threshold = "0.0",
  Saga.Segmentation.Refresh = "0", Saga.Segmentation.Leafsize = 256,
  Split = "0", Grass.Segmentation.Threshold = NULL,
  Grass.Segmentation.Weighted = FALSE,
  Grass.Segmentation.Method = "region_growing",
  Grass.Segmentation.Similarity = "euclidean",
  Grass.Segmentation.Minsize = 15, Grass.Segmentation.Memory = 300,
  Grass.Segmentation.Iterations = 50, Grass.Segmentation.Seeds = NULL,
  Grass.Segmentation.Neighbourhood = "0",
  Segmentation.Boundary.Grid = NULL,
  Grass.Segmentation.Goodness = paste0("Grass.Segmentation.Goodness",
  par.i), AllVertices = "FALSE", NoData = FALSE, Mask = NULL,
  NoData.Flag = -99999, show.output.on.console = FALSE,
  Seed.Method = "", Seed.Generation.Variance = paste0(tempdir(),
  paste0("SeedGenerationVariance", par.i, ".sgrd")),
  Seed.Generation.Points = file.path(tempdir(),
  paste0("SeedGenerationPoints", par.i, ".shp")),
  Seed.Generation.Type = "0", Seed.Generation.Scale = "10.0",
  Generalisation.Flac = FALSE, Generalization.Mode = "1",
  Generalization.Radius = "1", Generalization.Threshold = "0.0",
  env = RSAGA::rsaga.env(), Grass.SLIC.Iter = 10,
  Grass.SLIC.Superpixels = 200, Grass.SLIC.Step = 0,
  Grass.SLIC.Compactness = 1, Grass.SLIC.Superpixels.MinSize = 1,
  Grass.SLIC.Memory = 300, Grass.SLIC.Perturb = 0,
  burn.Boundary.into.Segments = FALSE, estimateScaleParameter = FALSE,
  Mode.Filter.Flac = FALSE, Mode.Filter.Size = 7,
  Mode.Filter.Segment.MinSize = 3, par.i = "", Sieving.Flac = FALSE,
  Sieving.Mode = "0", Sieving.Thresh = 4, Sieving.Expand = 4, ...)

Arguments

Tool	GRASS, SAGA or GRASS Superixels SLIC. Definition of open-source software which will be used
Segments.Grid	output path of raster with segments
Segments.Poly	output path of polygon with segments
Input.Grid	vector containing grid(s) for segmentation. It is possible to add multiple gridsm, as well as different grids even in combination of SAGA and GRASS. By using SAGA and GRASS combination the following separation must be used: '<>' (SAGA before GRASS grids!)
Saga.Output.Grid	output of FAST REPRESENTATIVENESS in SAGA. Default: temp
Saga.Output.Lod	output Lod of Representativeness Function in SAGA. Default: temp
Output.Seeds	output of seed points as raster, used for segmentation. Default: temp
Fast.Representativeness.LevelOfGeneralisation	determining number of seed points. Default: "10.0"
Saga.Similarity	output of similarity grid. Default: temp
Saga.Segments.Seeds.Table	table of seeds information. Default: temp
Saga.Segmentation.Normalize	normalisation during imagery segmentation. Default: "0"
Saga.Segmentation.Neighbourhood	neighbourhood considered during imagery segmentation. Default: "0" (4, alternative: "1" for 8)
Saga.Segmentation.Method	segmentation method during imagery segmentation. Default: "0" (feature space and position)
Saga.Segmentation.Sig.1	variance in feature space in imagery segmentation. Default: "1.0"
Saga.Segmentation.Sig.2	variance in position space in imagery segmentation. Default: "1.0"
Saga.Segmentation.Threshold	similarity threshold for joining pixel to segments in imagery segmentation. Default: "0.0"
Saga.Segmentation.Refresh	refresh image after imagery segmentation. Default: "0"
Saga.Segmentation.Leafsize	parameter for speed optimation in imagery segmentation. Default: 256
Split	split polygons to singlepart in vectorising grid classes. Default: "0"
Grass.Segmentation.Threshold	similarity threshold for joining pixel to segments. Default: NULL, 0.0 is not allowed
Grass.Segmentation.Weighted	option of weighing input grids in segmentation. Default: "FALSE"
Grass.Segmentation.Method	type of GRASS Segmentation. Default: "region_growing"
Grass.Segmentation.Similarity	distance measurement of similarity. Default: "euclidean"
Grass.Segmentation.Minsize	minsize of segment. Default: 15
Grass.Segmentation.Memory	memory to be used for segmentation. Default: 300
Grass.Segmentation.Iterations	amount of allowed iterations. Default: 50
Grass.Segmentation.Seeds	input of seeds raster. Enables bottom-up segmentation. Default: NULL
Segmentation.Boundary.Grid	input of boundary raster. Enables top-down (or hierarchical) segmentation. Default: NULL. NULL values must be 0 (or any other not segment value!).
Grass.Segmentation.Goodness	name for output goodness of fit estimate map. Default:"Grass.Segmentation.Goodness"
AllVertices	use all vertices by vectorising grid classes. Default: "FALSE"
NoData	input data contains NoData value. Default: FALSE
Mask	mask raster to mask NoData from input. Default: NULL
show.output.on.console	show output on console. Default: FALSE
Seed.Method	type of seed method for getting seeds. Default: "" (alternative: "Fast Representativeness", "Seed Generation", "Superpixels SLIC")
Seed.Generation.Variance	output raster with variance of seed generation. Default: temp
Seed.Generation.Points	output of seed points as shapefile. Default: temp
Seed.Generation.Type	option of seed generation type. Default:"0" (minima of variance, alternative: maxima of variance)
Seed.Generation.Scale	determining number of seed points in seed generation. Default: "10.0"
Generalisation.Flac	performing (multiple) majority filter on segments. Default: FALSE
Generalization.Mode	search mode by filtering: Default: "1" (circle, alternative: square)
Generalization.Threshold	threshold for applying majority filters. Default: "0.0"
env	environment of RSAGA. Default: RSAGA::rsaga.env()
Grass.SLIC.Iter	maximum number of iterations. Default: 10
Grass.SLIC.Superpixels	approximate number of output super pixels. Default: 200
Grass.SLIC.Step	distance (number of cells) between initial super pixel centers. A step size > 0 overrides the number of super pixels. Default: 0
Grass.SLIC.Compactness	compactness. A larger value causes more compact superpixels. Default: 1.0
Grass.SLIC.Superpixels.MinSize	minimum superpixel size. Default: 1
Grass.SLIC.Memory	memory in MB. Default: 300
Grass.SLIC.Perturb	Perturb initial super pixel centers. Percent of intitial superpixel radius. Default: 0, range: 0-100
burn.Boundary.into.Segments	vector specifing if boundary grid is burned into segmentation (1) or seeds (2). Default: FALSE, maximum length: 2
estimateScaleParameter	must be only be TRUE when scale parameter function is used. Default: FALSE
Mode.Filter.Flac	re-assign objects of a specific size based on mode values to its surroundings (moving window). Default: FALSE
Mode.Filter.Size	moving window size of mode-filter. Default: 3
Mode.Filter.Segment.MinSize	objects smaller and equal to this size are selected for filtering. Default: 3
par.i	run number. Default: ""
Sieving.Flac	perform sieving. Default: FALSE
Sieving.Mode	sieving mode. Default:"0"
Sieving.Thresh	minsize of clumps. Default: 4
Sieving.Expand	expand cells using majority filter. radius in cell sizes. Default:4

ggRaver/Lslide documentation built on April 8, 2022, 7:14 a.m.