Title:	Metrics for Assessing Segmentation Accuracy for Geospatial Data
Version:	0.3.0
Description:	A system that computes metrics to assess the segmentation accuracy of geospatial data. These metrics calculate the discrepancy between segmented and reference objects, and indicate the segmentation accuracy. For more details on choosing evaluation metrics, we suggest seeing Costa et al. (2018) <doi:10.1016/j.rse.2017.11.024> and Jozdani et al. (2020) <doi:10.1016/j.isprsjprs.2020.01.002>.
License:	MIT + file LICENSE
Encoding:	UTF-8
LazyData:	true
RoxygenNote:	7.2.3
Imports:	magrittr, graphics, units
Suggests:	classInt, dplyr, testthat (≥ 3.0.0)
Depends:	R (≥ 2.10), sf
BugReports:	https://github.com/michellepicoli/segmetric/issues
URL:	https://michellepicoli.github.io/segmetric/
NeedsCompilation:	no
Packaged:	2023-01-10 12:12:54 UTC; araujopicoli
Author:	Rolf Simoes [aut], Alber Sanchez [aut], Michelle Picoli [aut, cre]
Maintainer:	Michelle Picoli <mipicoli@gmail.com>
Repository:	CRAN
Date/Publication:	2023-01-10 12:30:02 UTC

segmetric

Description

Metrics for assessing segmentation accuracy for geospatial data.

Purpose

The segmetric package provides a set of metrics for the segmentation accuracy assessment (or evaluation) of geospatial data. It includes more than 20 metrics used in the literature for spatial segmentation assessment (Van Rijsbergen, 1979; Levine and Nazif, 1982; Janssen and Molenaar, 1995; Lucieer and Stein, 2002; Carleer et al., 2005; Moller et al., 2007; van Coillie et al., 2008; Costa et al., 2008; Weidner, 2008; Feitosa et al., 2010; Clinton et al. 2010; Persello and Bruzzone, 2010; Yang et al., 2014; and Zhang et al., 2015).

Extensions

The segmetric package is extensible and provides a set of functions to ease the implementation of new metrics. See ?sm_reg_metric() to find how new metrics are implemented.

Contributions

Contribution to this package could be done at segmetric's page on GitHub: https://github.com/michellepicoli/segmetric.

Author(s)

Maintainer: Michelle Picoli mipicoli@gmail.com (ORCID)

Authors:

• Rolf Simoes rolfsimoes@gmail.com (ORCID)

• Alber Sanchez albhasan@gmail.com (ORCID)

References

• Carleer, A.P., Debeir, O., Wolff, E., 2005. Assessment of very high spatial resolution satellite image segmentations. Photogramm. Eng. Remote. Sens. 71, 1285-1294. doi: 10.14358/PERS.71.11.1285.

• Clinton, N., Holt, A., Scarborough, J., Yan, L., Gong, P., 2010. Accuracy assessment measures for object-based image segmentation goodness. Photogramm. Eng. Remote. Sens. 76, pp. 289-299.

• Costa, G.A.O.P., Feitosa, R.Q., Cazes, T.B., Feijo, B., 2008. Genetic adaptation of segmentation parameters. In: Blaschke, T., Lang, S., Hay, G.J. (Eds.), Object-based Image Analysis. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 679-695. doi: 10.1007/978-3-540-77058-9_37.

• Dice, L.R., 1945. Measures of the amount of ecologic association between species. Ecology, 26(3), pp.297-302.

• Feitosa, R.Q., Ferreira, R.S., Almeida, C.M., Camargo, F.F., Costa, G.A.O.P., 2010. Similarity metrics for genetic adaptation of segmentation parameters. In: 3rd International Conference on Geographic Object-Based Image Analysis (GEOBIA 2010). The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Ghent.

• Jaccard, P., 1912. The distribution of the flora in the alpine zone.

1. New phytologist, 11(2), pp.37-50. doi: 10.1111/j.1469-8137.1912.tb05611.x

• Janssen, L.L.F., Molenaar, M., 1995. Terrain objects, their dynamics and their monitoring by the integration of GIS and remote sensing. IEEE Trans. Geosci. Remote Sens. 33, pp. 749-758. doi: 10.1109/36.387590.

• Levine, M.D., Nazif, A.M., 1982. An experimental rule based system for testing low level segmentation strategies. In: Preston, K., Uhr, L. (Eds.), Multicomputers and Image Processing: Algorithms and Programs. Academic Press, New York, pp. 149-160.

• Lucieer, A., Stein, A., 2002. Existential uncertainty of spatial objects segmented from satellite sensor imagery. Geosci. Remote. Sens. IEEE Trans. 40, pp. 2518-2521. doi: 10.1109/TGRS.2002.805072.

• Möller, M., Lymburner, L., Volk, M., 2007. The comparison index: a tool for assessing the accuracy of image segmentation. Int. J. Appl. Earth Obs. Geoinf. 9, pp. 311-321. doi: 10.1016/j.jag.2006.10.002.

• Persello, C., Bruzzone, L., 2010. A novel protocol for accuracy assessment in classification of very high resolution images. IEEE Trans. Geosci. Remote Sens. 48, pp. 1232-1244. doi: 10.1109/TGRS.2009.2029570.

• Rezatofighi, H., Tsoi, N., Gwak, J., Sadeghian, A., Reid, I., Savarese, S.,

1. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 658-666.

• Van Coillie, F.M.B., Verbeke, L.P.C., De Wulf, R.R., 2008. Semi-automated forest stand delineation using wavelet based segmentation of very high resolution optical imagery. In: Object-Based Image Analysis: Spatial Concepts for Knowledge-Driven Remote Sensing Applications, pp. 237-256. doi: 10.1007/978-3-540-77058-9_13.

• Van Rijsbergen, C.J., 1979. Information Retrieval. Butterworth-Heinemann, London.

• Weidner, U., 2008. Contribution to the assessment of segmentation quality for remote sensing applications. In: Proceedings of the 21st Congress for the International Society for Photogrammetry and Remote Sensing, 03–11 July, Beijing, China. Vol. XXXVII. Part B7, pp. 479-484.

• Yang, J., Li, P., He, Y., 2014. A multi-band approach to unsupervised scale parameter selection for multi-scale image segmentation. ISPRS J. Photogramm. Remote Sens. 94, pp. 13-24. doi: 10.1016/j.isprsjprs.2014.04.008.

• Yang, J., He, Y., Caspersen, J. P., Jones, T. A., 2017. Delineating Individual Tree Crowns in an Uneven-Aged, Mixed Broadleaf Forest Using Multispectral Watershed Segmentation and Multiscale Fitting. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 10(4), pp. 1390-1401. doi: 10.1109/JSTARS.2016.2638822.

• Zhan, Q., Molenaar, M., Tempfli, K., Shi, W., 2005. Quality assessment for geo‐spatial objects derived from remotely sensed data. International Journal of Remote Sensing, 26(14), pp.2953-2974. doi: 10.1080/01431160500057764

• Zhang, X., Feng, X., Xiao, P., He, G., Zhu, L., 2015a. Segmentation quality evaluation using region-based precision and recall measures for remote sensing images. ISPRS J. Photogramm. Remote Sens. 102, pp. 73-84. doi: 10.1016/j.isprsjprs.2015.01.009.

See Also

Useful links:

• https://michellepicoli.github.io/segmetric/

• Report bugs at https://github.com/michellepicoli/segmetric/issues

Database of metrics functions

Description

These functions are used to register new metrics in the segmetric database.

• sm_list_metrics(): List all registered metrics.

• sm_new_metric(): Create a new metric entry to be registered in the database.

• sm_reg_metric(): Register a new metric in the database.

• sm_unreg_metric(): Remove a metric entry from the database.

• sm_desc_metric(): Describe a metric registered from the database.

Usage

.db_list()

.db_set(key, value)

.db_del(key)

.db_get(key)

sm_list_metrics()

sm_new_metric(
  fn,
  fn_subset,
  name = "",
  optimal = 0,
  summarizable = TRUE,
  description = "",
  reference = ""
)

sm_reg_metric(metric_id, entry)

sm_unreg_metric(metric_id)

sm_desc_metric(metric_id)

Arguments

Value

• sm_list_metrics(): Return character vector with supported metrics.

• sm_new_metric(): Return a metric_entry object containing the metric function (fn), name (name), description (description), and reference (reference).

• sm_reg_metric(), sm_unreg_metric(): No return value, called to (un)register a metric.

• sm_desc_metric(): No return value, called to print a metric description.

Examples

sm_reg_metric(
    metric_id = "Example",
    entry = sm_new_metric(
        fn = function(m, ...) {
            sm_area(sm_ytilde(m)) / 
                sm_area(sm_ref(m), order = sm_ytilde(m))
        },
        fn_subset = sm_ytilde,
        name = "Metric name example",
        description = paste("Values range from A to B.",
                            "Optimal value is C"),
        reference = "Author (Year)"
    ))

sm_desc_metric("Example")
sm_list_metrics()

General functions

Description

These functions manipulate segmetric objects.

• sm_area(): Return a vector of areas, one for each polygon.

• sm_centroid(): Return the centroids of the given polygons.

• sm_intersection(): Return the intersection of the given simple features.

• sm_subset_union(): Return the union of the given simple features.

• sm_rbind(): Return the merge of unique simple features.

• sm_summarize_groups(): Summarizes metric values by a group (either reference or segment).

• sm_options(): Get/Set segmetric options.

Usage

sm_area(s, order = NULL)

sm_centroid(s, order = NULL)

sm_distance(s1, s2)

sm_intersection(s1, s2, touches = TRUE)

sm_subset_union(s)

sm_rbind(...)

sm_apply_group(x, groups, fn, ...)

sm_summarize_group(x, groups, fn, ...)

sm_norm_left(x, y)

sm_norm_right(x, y)

sm_options(..., digits = NULL)

Arguments

Value

• sm_area(): Return a numeric vector with polygons' area.

• sm_centroid(): Return a subset_sf object with polygons' centroid.

• sm_intersection(): Return a subset_sf object with intersection between polygons.

• sm_subset_union(): Return a subset_sf object with union between intersecting polygons.

• sm_rbind(): Return a subset_sf object with unique features.

Metric functions

Description

The sm_compute() computes a given metric (metric_id parameter) from segmentation objects. It compares the reference to the segmentation polygons using a metric.

A list with all supported metrics can be obtained by sm_list_metrics() (see Details for more information).

The sm_metric_subset() returns the subset used to compute the metrics in segmetric object.

Usage

sm_compute(m, metric_id, ...)

sm_metric_subset(m, metric_id = NULL)

Arguments

Details

• "OS1" refers to Oversegmentation. Its values range from 0 (optimal) to 1 (Clinton et al., 2010).

• "US1" refers to Undersegmentation. Its values range from 0 (optimal) to 1 (Clinton et al., 2010).

• "OS2" refers to Oversegmentation. Its values range from 0 (optimal) to 1 (Persello and Bruzzone, 2010).

• "US2" refers to Undersegmentation. Its values range from 0 (optimal) to 1 (Persello and Bruzzone, 2010).

• "OS3" refers to Oversegmentation. Its values range from 0 (optimal) to 1 (Yang et al., 2014).

• "US3" refers to Undersegmentation. Its values range from 0 (optimal) to 1 (Yang et al., 2014).

• "AFI" refers to Area Fit Index. Its optimal value is 0 (Lucieer and Stein, 2002; Clinton et al., 2010).

• "QR" refers to Quality Rate. Its values range from 0 (optimal) to 1 (Weidner, 2008; Clinton et al., 2010).

• "D_index" refers to Index D. Its values range from 0 (optimal) to 1 (Levine and Nazif, 1982; Clinton et al., 2010).

• "precision" refers to Precision. Its values range from 0 to 1 (optimal) (Van Rijsbergen, 1979; Zhang et al., 2015).

• "recall" refers to Recall. Its values range from 0 to 1 (optimal) (Van Rijsbergen, 1979; Zhang et al., 2015).

• "UMerging" refers to Undermerging. Its values range from 0 (optimal) to 1 (Levine and Nazif, 1982; Clinton et al., 2010).

• "OMerging" refers to Overmerging. Its optimal value is 0 (Levine and Nazif, 1982; Clinton et al., 2010).

• "M" refers to Match. Its values range from 0 to 1 (optimal) (Janssen and Molenaar, 1995; Feitosa et al., 2010).

• "E" refers to Evaluation Measure. Its values range from 0 (optimal) to 100 (Carleer et al., 2005).

• "RAsub" refers to Relative Area. Its values range from 0 to 1 (optimal) (Müller et al., 2007; Clinton et al., 2010).

• "RAsuper" refers to Relative area. Its values range from 0 to 1 (optimal) (Müller et al., 2007; Clinton et al., 2010).

• "PI" refers to Purity Index. Its values range from 0 to 1 (optimal) (van Coillie et al., 2008).

• "Fitness" refers to Fitness Function. Its optimal value is 0 (Costa et al., 2008).

• "ED3" refers to Euclidean Distance. Its values range from 0 (optimal) to 1 (Yang et al., 2014).

• "F_measure" refers to F-measure metric. Its values range from 0 to 1 (optimal) (Van Rijsbergen, 1979; Zhang et al., 2015). It takes the optional weight argument alpha, ranging from 0.0 to 1.0 (the default is 0.5).

• "IoU" refers to Intersection over Union metric. Its values range from 0 to 1 (optimal) (Jaccard, 1912; Rezatofighi et al., 2019).

• "SimSize" refers to the similarity size metric. Its values range from 0 to 1 (optimal) (Zhan et al., 2005).

• "qLoc"refers to quality of object’s location metric. Its optimal value is 0 (Zhan et al., 2005).

• "RPsub" refers to Relative Position (sub) metric. Optimal value is 0 (Möller et al., 2007, Clinton et al., 2010).

• "RPsuper" refers to Relative Position (super) metric. Its values range from 0 (optimal) to 1 (Möller et al., 2007, Clinton et al., 2010).

• "OI2 refers to Overlap Index metric. Its values range from 0 to 1 (optimal) (Yang et al., 2017).

Value

Return a numeric vector with computed metric.

References

A complete list of cited references is available in ?segmetric.

See Also

sm_list_metrics()

Examples

# load sample datasets
data("sample_ref_sf", package = "segmetric")
data("sample_seg_sf", package = "segmetric")

# create segmetric object
m <- sm_read(ref_sf = sample_ref_sf, seg_sf = sample_seg_sf)

# compute AFI metric and summarize it
sm_compute(m, "AFI") %>% summary()

# compute three metrics and summarize them
sm_compute(m, c("AFI", "OS1", "US2")) %>% summary()

# compute OS1, F_measure, and US2 metrics using pipe
m <- sm_compute(m, "OS1") %>%
  sm_compute("F_measure") %>%
  sm_compute("US2")

# summarize them
summary(m)

Plot function

Description

Plot a segmetric map according to the parameter type:

• "base": simple plot of the reference or segmentation polygons;

• "subset": plot polygons from a subset over the base plot;

• "choropleth": plot a choropleth map from polygons of a subset using metric values.

Usage

## S3 method for class 'segmetric'
plot(
  x,
  type = "base",
  ...,
  title = NULL,
  layers = c("ref_sf", "seg_sf"),
  background = "#FFFFFF",
  ref_color = "#FF00009F",
  ref_fill = "#FFFFFF00",
  ref_label = "reference",
  ref_size = 2,
  ref_symbol = 2,
  seg_color = "#0000009F",
  seg_fill = "#FFFFFF00",
  seg_label = "segment",
  seg_size = 1,
  seg_symbol = 3,
  selected_fill = "#9A9AFF50",
  plot_centroids = TRUE,
  centroids_color = "#000000FF",
  centroids_label = "centroid",
  subset_id = NULL,
  subset_color = "#FFFFFF00",
  subset_fill = "#F0E4167F",
  metric_id = NULL,
  break_style = "jenks",
  choropleth_palette = "YlGnBu",
  choropleth_palette_reverse = FALSE,
  choropleth_size = 0.1,
  plot_extent = NULL,
  plot_legend = TRUE,
  plot_axes = TRUE
)

Arguments

Objects exported from other packages

Description

These objects are imported from other packages. Follow the links below to see their documentation.

magrittr

%>%

LEM+ dataset

Description

ref_sf: a dataset containing field boundaries from Luiz Eduardo Magalhaes municipality, Brazil.

The data covers the following extent: xmin: -46.37683 ymin: -12.34579 xmax: -46.15776 ymax: -12.13663 CRS: EPSG:4326

sample_ref_sf: a subset of ref_sf dataset.

Usage

ref_sf

sample_ref_sf

Format

These datasets are objects of class sf (inherited from tbl_df, tbl, data.frame) with 2 variables:

• id: identification

• geometry: polygons

ref_sf: a dataset with 195 features.

sample_ref_sf: a dataset with 5 features.

Source

Oldoni et al. (2020) doi: 10.1016/j.dib.2020.106553.

References

• Oldoni, L.V., Sanches, I.D.A., Picoli, M.C.A., Covre, R.M. and Fronza, J.G., 2020. LEM+ dataset: For agricultural remote sensing applications. Data in Brief, 33, p.106553.

Examples

data("ref_sf", package = "segmetric")
data("sample_ref_sf", package = "segmetric")

Segmentation dataset

Description

seg200_sf,seg500_sf,seg800_sf,seg1000_sf: a dataset containing segments generated from PlanetScope image, level 3B, acquired on Feb 18, 2020, with 3.7-meter resolution (Planet Team, 2017), using the multiresolution segmentation method (Baatz and Schape, 2000).

The data covers the approximately the same area of LEM+ dataset (see ref_sf).

The data was post-processed using the spectral difference algorithm on band 3.

The polygons were simplified using the Douglas-Peucker algorithm in QGIS.

Self-intersections were removed using SAGA's Polygon Self-Intersection.

Segmentation parameters:

• ⁠scale parameter⁠: 200 (seg200_sf), 500 (seg500_sf), 800 (seg800_sf), and 1000 (seg1000_sf)

• shape: 0.9

• compactness: 0.1

Spectral difference parameters:

• ⁠spectral difference⁠: 20

Simplification parameter:

• distance: 10-meters

Only those polygons intersecting reference data with an area-perimeter ratio above 25 were selected.

sample_seg_sf: a subset of seg_sf dataset.

Usage

seg200_sf

seg500_sf

seg800_sf

seg1000_sf

sample_seg_sf

Format

These datasets are objects of class sf (inherited from tbl_df, tbl, data.frame) with 2 variables:

• id: identification

• geometry: polygons

seg200_sf: a dataset with 547 features. seg500_sf: a dataset with 215 features. seg800_sf: a dataset with 169 features. seg1000_sf: a dataset with 158 features.

An object of class sf (inherits from tbl_df, tbl, data.frame) with 215 rows and 2 columns.

An object of class sf (inherits from tbl_df, tbl, data.frame) with 169 rows and 2 columns.

An object of class sf (inherits from tbl_df, tbl, data.frame) with 158 rows and 2 columns.

sample_seg_sf: a dataset with 6 features extracted from seg500_sf dataset.

References

• Planet Team, 2017. Planet Application Program Interface: In Space for Life on Earth. San Francisco, CA. https://www.planet.com

• Baatz, M., Schape, A., 2000. Multiresolution segmentation - an optimization approach for high quality multi-scale image segmentation. In: Strobl, J., Blaschke, T., Griesebner, G. (Eds.), Angewandte Geographische Informations-Verarbeitung XII. Wichmann Verlag, Karlsruhe, Germany, pp. 12-23. <>

Examples

data("seg200_sf", package = "segmetric")
data("seg500_sf", package = "segmetric")
data("seg800_sf", package = "segmetric")
data("seg1000_sf", package = "segmetric")
data("sample_seg_sf", package = "segmetric")

General functions

Description

These functions manipulate segmetric objects.

• sm_read(): Load the reference and segmentation polygons into segmetric.

• sm_clear(): Remove the already calculated metrics from segmetric.

• print(): Print a segmetric object.

• plot(): Plot the reference and segmentation polygons.

• summary(): Compute a measure of central tendency over the values of a metric.

• sm_is_empty(): Check if a segmetric object is empty.

Usage

.segmetric_check(m)

.segmetric_env(m)

sm_read(ref_sf, seg_sf)

sm_clear(m)

## S3 method for class 'segmetric'
summary(object, weight = NULL, na_rm = TRUE, ...)

sm_is_empty(m)

## S3 method for class 'segmetric'
x[i]

Arguments

Value

• sm_read(), sm_clear(): Return a segmetric object containing an empty list and an environment attribute to store the necessary datasets.

• sm_is_empty(): Return a logical vector indicating if each computed metric is empty.

See Also

sm_compute()

Examples

# load sample datasets
data("sample_ref_sf", package = "segmetric")
data("sample_seg_sf", package = "segmetric")

# create segmetric object
m <- sm_read(ref_sf = sample_ref_sf, seg_sf = sample_seg_sf)

# plot geometries
plot(m)

# compute a metric
sm_compute(m, "AFI")

# summarize the metric using mean
sm_compute(m, "AFI") %>% summary()

# clear computed subsets
sm_clear(m)

Set functions

Description

These functions compute subsets required to calculate segmentation metrics as described in Clinton et al. (2010) and Costa et al. (2017).

• sm_ref() returns the set of n polygons of reference, represented by X = \{x_{i}: i = 1, ....., n\}

• sm_seg() returns the set of m segmentation polygons, represented by Y = \{y_{j}: j = 1, ....., m\}

• sm_ytilde() returns \tilde{Y}_{i}, a subset of Y, where \tilde{Y}_{i} = \{y_{j}: \rm{area}(x_{i} \cap y_{j}) \neq 0\}

• sm_xtilde() returns \tilde{X}_{j}, a subset of X, where \tilde{X}_{j} = \{x_{i}: \rm{area}(y_{j} \cap x_{i}) \neq 0\}

• sm_yprime() returns Y'_{i}, a subset of Y, where Y'_{i} = \{y_{j}: max(\rm{area}(x_{i} \cap y_{j}))\}

• sm_xprime() returns X'_{j}, a subset of X, where X'_{j} = \{x_{i}: max(\rm{area}(y_{j} \cap x_{i}))\}

• sm_ya() returns Y\!a_{i}, a subset of \tilde{Y}_{i}, where Y\!a_{i} = \{y_{j}: \rm{centroid}(x_{i}) \:\rm{in}\: y_{j}\}

• sm_yb() returns Y\!b_{i}, a subset of \tilde{Y}_{i}, where Y\!b_{i} = \{y_{j}: \rm{centroid}(y_{j}) \:\rm{in}\: x_{i}\}

• sm_yc() returns Y\!c_{i}, a subset of \tilde{Y}_{i}, where Y\!c_{i} = \{y_{j}: \rm{area}(x_{i} \cap y_{j}) / \rm{area}(y_{j}) > 0.5\}

• sm_yd() returns Y\!d_{i}, a subset of \tilde{Y}_{i}, where Y\!d_{i} = \{y_{j}: \rm{area}(x_{i} \cap y_{j}) / \rm{area}(x_{i}) > 0.5\}

• sm_ystar() returns {Y}^{*}_{i}, where {Y}^{*}_{i} = Y\!a_{i} \cup Y\!b_{i} \cup Y\!c_{i} \cup Y\!c_{i}

• sm_ycd() returns Y\!cd_{i}, where Y\!cd_{i} = Y\!c_{i} \cup Y\!d_{i}

• sm_ye() returns Y\!e_{i}, a subset of \tilde{Y}_{i}, where Y\!e_{i} = \{y_{j}: \rm{area}(x_{i} \cap y_{j}) / \rm{area}(y_{j}) = 1\}

• sm_yf() returns Y\!f_{i}, a subset of \tilde{Y}_{i}, where Y\!f_{i} = \{y_{j}: \rm{area}(x_{i} \cap y_{j}) / \rm{area}(y_{j}) > 0.55\}

• sm_yg() returns Y\!g_{i}, a subset of \tilde{Y}_{i}, where Y\!g_{i} = \{y_{j}: \rm{area}(x_{i} \cap y_{j}) / \rm{area}(y_{j}) > 0.75\}

Usage

sm_ytilde(m)

sm_xtilde(m)

sm_yprime(m)

sm_xprime(m)

sm_ya(m)

sm_yb(m)

sm_yc(m)

sm_yd(m)

sm_ystar(m)

sm_ycd(m)

sm_ye(m)

sm_yf(m)

sm_yg(m)

Arguments

Value

• sm_ref(): Return an object of class ref_sf (inherited from sf) containing identification (ref_id) and geometry (geometry) columns.

• sm_seg(): Return an object of class seg_sf (inherited from sf) containing identification (seg_id) and geometry (geometry) columns.

• sm_ytilde(), sm_xtilde(), sm_yprime(), sm_xprime(), sm_ya(), sm_yb(), sm_yc(), sm_yd(), sm_ystar(), sm_ycd(), sm_ye(), sm_yf(), and sm_yg(): Return an object of class subset_sf (inherited from sf) containing identification (ref_id and seg_id), and geometry (geometry) columns.

References

Clinton, N., Holt, A., Scarborough, J., Yan, L., & Gong, P. (2010). Accuracy Assessment Measures for Object-based Image Segmentation Goodness. Photogrammetric Engineering & Remote Sensing, 76(3), 289–299. doi: 10.14358/PERS.76.3.289.

Costa, H., Foody, G. M., & Boyd, D. S. (2018). Supervised methods of image segmentation accuracy assessment in land cover mapping. Remote Sensing of Environment, 205(December 2017), 338–351. doi: 10.1016/j.rse.2017.11.024.

Subset handling functions

Description

These functions are intended to be used in new metric extensions. They handle subset_sf data (inherited from sf class) stored in segmetric objects.

• sm_list() lists subsets already computed and stored in a segmetric object.

• sm_exists() verifies if a subset_id exists in a segmetric object.

• sm_subset() evaluates and stores a subset_sf object.

• sm_indirect() finds the subset_id of a given subset_sf object stored in a segmetric object.

• sm_segmetric() returns the segmetric object that stores a given subset object (either a ref_sf, a seg_sf, or a subset_sf).

• sm_get() retrieves a subset_sf object stored in a segmetric object.

• sm_inset() operator equivalent to inner join but returns only objects from s1, or its corresponding row in s2 if parameter return_index is TRUE.

• sm_group_by(): Apply a function to groups of subset_sf.

Usage

sm_list(m)

sm_exists(m, subset_id)

sm_subset(m, subset_id, expr = NULL)

sm_indirect(s)

sm_segmetric(s)

sm_get(m, subset_id)

sm_ref(m)

sm_seg(m)

sm_inset(s1, s2, return_index = FALSE)

## S3 method for class 'ref_sf'
sm_inset(s1, s2, return_index = FALSE)

## S3 method for class 'seg_sf'
sm_inset(s1, s2, return_index = FALSE)

## S3 method for class 'subset_sf'
sm_inset(s1, s2, return_index = FALSE)

sm_group_by(s, by, fn, ...)

Arguments

Value

• sm_list(): Return a character vector with all names of subsets stored in the segmetric object.

• sm_exists(): Return a logical value indicating if a given subset name is stored in the segmetric object.

• sm_subset(): Return a subset_sf object.

• sm_indirect(): Return the subset name of a given subset_sf object stored in a segmetric object.

• sm_segmetric(): Return a segmetric object that stores a given subset_sf object.

• sm_get(): Return a subset_sf object stored in a segmetric object.

• sm_inset(): Return either a subset_sf object or an integer vector with the index of corresponding rows of s2 object.

• sm_group_by(): Return a subset_sf object.

Examples

# load sample datasets
data("sample_ref_sf", package = "segmetric")
data("sample_seg_sf", package = "segmetric")

# create segmetric object
m <- sm_read(ref_sf = sample_ref_sf, seg_sf = sample_seg_sf)

# lists only 'ref_sf' and 'seg_sf'
sm_list(m)

# computes 'Y_tilde' subset and stores it as 'test_subset' subset id
# sm_ytilde(m) also stores a subset under 'Y_tilde' id
s <- sm_subset(m, "test_subset", sm_ytilde(m))

# lists 'ref_sf', 'seg_sf', 'test_subset', and 'Y_tilde'
sm_list(m)

# which segmetric object stores `s` subset?
m2 <- sm_segmetric(s)

# m is identical to m2
identical(m, m2)

# which name `s` subset is stored in `m` segmetric object?
sm_indirect(s)

# retrieve 'test_subset' data from `m` object
s2 <- sm_get(m, 'test_subset')

# s is identical to s2
identical(s, s2)