| Title: | Classify Open Street Map Features |
|---|---|
| Description: | Classify Open Street Map (OSM) features into meaningful functional or analytical categories. Designed for OSM PBF files, e.g. from <https://download.geofabrik.de/> imported as spatial data frames. A classification consists of a list of categories that are related to certain OSM tags and values. Given a layer from an OSM PBF file and a classification, the main osm_classify() function returns a classification data table giving, for each feature, the primary and alternative categories (if there is overlap) assigned, and the tag(s) and value(s) matched on. The package also contains a classification of OSM features by economic function/significance, following Krantz (2023) <https://www.ssrn.com/abstract=4537867>. |
| Authors: | Sebastian Krantz [aut, cre] |
| Maintainer: | Sebastian Krantz <[email protected]> |
| License: | GPL-3 |
| Version: | 0.1.3 |
| Built: | 2024-10-31 19:45:55 UTC |
| Source: | https://github.com/sebkrantz/osmclass |
An R package to classify Open Street Map (OSM) features into meaningful functional or analytical categories.
It expects OSM PBF data, e.g. from https://download.geofabrik.de/, imported as data frames (e.g. using sf), and
is well optimized to deal with large quantities of OSM data.
Main Function to Classify OSM Features
Auxiliary Functions to Extract Information (Tags) from OSM PBF Layers
osm_other_tags_list()osm_tags_df()
A Classification of OSM Features by Economic Function, developed for the Africa OSM following Krantz (2023)
osm_point_polygon_classosm_line_classosm_line_info_tags
Krantz, Sebastian, Mapping Africa’s Infrastructure Potential with Geospatial Big Data, Causal ML, and XAI (August 10, 2023). Available at SSRN: https://www.ssrn.com/abstract=4537867
## Not run: # Download OSM PBF file for Djibouti download.file("https://download.geofabrik.de/africa/djibouti-latest.osm.pbf", destfile = "djibouti-latest.osm.pbf", mode = "wb") # Import OSM data for Djibouti library(sf) st_layers("djibouti-latest.osm.pbf") points <- st_read("djibouti-latest.osm.pbf", "points") lines <- st_read("djibouti-latest.osm.pbf", "lines") polygons <- st_read("djibouti-latest.osm.pbf", "multipolygons") # Classify features library(osmclass) points_class <- osm_classify(points, osm_point_polygon_class) polygons_class <- osm_classify(polygons, osm_point_polygon_class) lines_class <- osm_classify(lines, osm_line_class) # See what proportion of the data we have classified sum(points_class$classified)/nrow(points) sum(polygons_class$classified)/nrow(polygons) sum(lines_class$classified)/nrow(lines) # Get some additional info for lines library(collapse) lines_info <- lines |> ss(lines_class$classified) |> rsplit(lines_class$main_cat[lines_class$classified]) |> get_vars(names(osm_line_info_tags), regex = TRUE) lines_info <- Map(osm_tags_df, lines_info, osm_line_info_tags[names(lines_info)]) str(lines_info) # Get 'other_tags' of points layer as list other_point_tags <- osm_other_tags_list(points$other_tags, values = TRUE) str(other_point_tags) # TIP: For larger OSM files, importing layers (esp. lines and polygons) at once # may not be feasible memory-wise. In this case, translating to GPKG and using # an SQL query for stepwise processing is helpful: library(fastverse) library(sf) # Get all Africa OSM (6 Gb) opt <- options(timeout = 6000) download.file("https://download.geofabrik.de/africa-latest.osm.pbf", destfile = "africa-latest.osm.pbf", mode = "wb") # GPKG is large (> 40 Gb) gdal_utils("vectortranslate", "africa-latest.osm.pbf", "africa-latest.gpkg") # Get map layers: shows how many features per layer layers <- st_layers("africa-latest.gpkg") print(layers) # Example: stepwise classifying lines, 1M features at a time N <- layers$features[layers$name == "lines"] int <- seq(0L, N, 1e6L) lines_class <- vector("list", length(int)) for (i in seq_len(length(int))) { cat("\nReading Lines Chunk:", i, "\n") temp = st_read("africa-latest.gpkg", query = paste("SELECT * FROM lines LIMIT 1000000 OFFSET", int[i])) # Some pre-selection: removing residential roads temp %<>% fsubset(is.na(highway) | highway %chin% osm_line_class$road$highway) # Classifying temp_class <- osm_classify(temp, osm_line_class) lines_class[[i]] <- ss(temp_class, temp_class$classified, check = FALSE) } # Combining lines_class <- rbindlist(lines_class) options(opt) ## End(Not run)## Not run: # Download OSM PBF file for Djibouti download.file("https://download.geofabrik.de/africa/djibouti-latest.osm.pbf", destfile = "djibouti-latest.osm.pbf", mode = "wb") # Import OSM data for Djibouti library(sf) st_layers("djibouti-latest.osm.pbf") points <- st_read("djibouti-latest.osm.pbf", "points") lines <- st_read("djibouti-latest.osm.pbf", "lines") polygons <- st_read("djibouti-latest.osm.pbf", "multipolygons") # Classify features library(osmclass) points_class <- osm_classify(points, osm_point_polygon_class) polygons_class <- osm_classify(polygons, osm_point_polygon_class) lines_class <- osm_classify(lines, osm_line_class) # See what proportion of the data we have classified sum(points_class$classified)/nrow(points) sum(polygons_class$classified)/nrow(polygons) sum(lines_class$classified)/nrow(lines) # Get some additional info for lines library(collapse) lines_info <- lines |> ss(lines_class$classified) |> rsplit(lines_class$main_cat[lines_class$classified]) |> get_vars(names(osm_line_info_tags), regex = TRUE) lines_info <- Map(osm_tags_df, lines_info, osm_line_info_tags[names(lines_info)]) str(lines_info) # Get 'other_tags' of points layer as list other_point_tags <- osm_other_tags_list(points$other_tags, values = TRUE) str(other_point_tags) # TIP: For larger OSM files, importing layers (esp. lines and polygons) at once # may not be feasible memory-wise. In this case, translating to GPKG and using # an SQL query for stepwise processing is helpful: library(fastverse) library(sf) # Get all Africa OSM (6 Gb) opt <- options(timeout = 6000) download.file("https://download.geofabrik.de/africa-latest.osm.pbf", destfile = "africa-latest.osm.pbf", mode = "wb") # GPKG is large (> 40 Gb) gdal_utils("vectortranslate", "africa-latest.osm.pbf", "africa-latest.gpkg") # Get map layers: shows how many features per layer layers <- st_layers("africa-latest.gpkg") print(layers) # Example: stepwise classifying lines, 1M features at a time N <- layers$features[layers$name == "lines"] int <- seq(0L, N, 1e6L) lines_class <- vector("list", length(int)) for (i in seq_len(length(int))) { cat("\nReading Lines Chunk:", i, "\n") temp = st_read("africa-latest.gpkg", query = paste("SELECT * FROM lines LIMIT 1000000 OFFSET", int[i])) # Some pre-selection: removing residential roads temp %<>% fsubset(is.na(highway) | highway %chin% osm_line_class$road$highway) # Classifying temp_class <- osm_classify(temp, osm_line_class) lines_class[[i]] <- ss(temp_class, temp_class$classified, check = FALSE) } # Combining lines_class <- rbindlist(lines_class) options(opt) ## End(Not run)
This classification, developed for Krantz (2023), aims to classify OSM features into meaningful and specific economic categories such as 'education', 'health', 'tourism', 'financial', 'shopping', 'transport', 'communications', 'industrial', 'residential', 'road', 'railway', 'pipeline', 'power', 'waterway' etc. Separate classifications are developed for points and polygons (buildings) (33 categories), and lines (11 categories), which should be applied to the respective layers of OSM PBF files, see osmclass-package for and example. The classification is optimized (in terms of tag choice and order of categories) to assign the most sensible primary category to most features in the Africa OSM.
osm_point_polygon_class osm_line_class osm_line_info_tagsosm_point_polygon_class osm_line_class osm_line_info_tags
An object of class list of length 33.
An object of class list of length 11.
An object of class list of length 11.
Krantz, Sebastian, Mapping Africa’s Infrastructure Potential with Geospatial Big Data, Causal ML, and XAI (August 10, 2023). Available at SSRN: https://www.ssrn.com/abstract=4537867
collapse::unlist2d(osm_point_polygon_class, idcols = c("category", "tag")) collapse::unlist2d(osm_line_class, idcols = c("category", "tag")) # This list contains additional tags with information about lines (e.g. roads and railways) collapse::unlist2d(osm_line_info_tags, idcols = c("category", "tag"))collapse::unlist2d(osm_point_polygon_class, idcols = c("category", "tag")) collapse::unlist2d(osm_line_class, idcols = c("category", "tag")) # This list contains additional tags with information about lines (e.g. roads and railways) collapse::unlist2d(osm_line_info_tags, idcols = c("category", "tag"))
A data table of all 8608 OSM points in Djibouti as of August 2023.
djibouti_pointsdjibouti_points
A data table with 8608 rows and 10 columns. The first column contains the OSM id of each point. Other columns give the values of frequent OSM tags for point features. The last column is called 'other_tags' and contains all remaining (less frequent) tags. Please consult the OSM Feature Documentation for the exact meaning and frequently used values of these tags.
Geofabrik download server (https://download.geofabrik.de/). See osmclass-package for how to download it.
data(djibouti_points)data(djibouti_points)
Classifies OSM features into meaningful functional or analytical categories, according to a supplied classification.
osm_classify(data, classification)osm_classify(data, classification)
data |
imported layer from an OSM PBF file. Usually an 'sf' data frame, but the geometry column is unnecessary. Importantly, the data frame should have an 'other_tags' column with OSM PBF formatting. |
|||||||||
classification |
a 2-level nested list providing a classification. The layers of the list are:
See |
a data.table with rows matching the input frame and columns
classified |
logical. Whether the feature was classified i.e. matched by any tag-value in the |
main_cat |
character. The first category the feature was assigned to, depending on the order of categories in the |
main_tag |
character. The tag matched for the main category. |
main_tag_value |
character. The value matched on. |
alt_cats |
character. Alternative (secondary) categories assigned, comma-separated if multiple. |
alt_tags_values |
character. The tags and double-quoted values matched for secondary categories, comma-separated if multiple. |
It is not necessary to expand the 'other_tags' column, e.g. using osm_tags_df(). osm_classify() efficiently searches the content of that column without expanding it.
# See Examples at ?osmclass for a full examples # Classify OSM Points in Djibouti djibouti_points_class <- osm_classify(djibouti_points, osm_point_polygon_class) head(djibouti_points_class) collapse::descr(djibouti_points_class)# See Examples at ?osmclass for a full examples # Classify OSM Points in Djibouti djibouti_points_class <- osm_classify(djibouti_points, osm_point_polygon_class) head(djibouti_points_class) collapse::descr(djibouti_points_class)
Generate a List from the 'other_tags' Column in OSM PBF Data
osm_other_tags_list(x, values = FALSE, split = "\",\"|\"=>\"", ...)osm_other_tags_list(x, values = FALSE, split = "\",\"|\"=>\"", ...)
x |
character. The 'other_tags' column of an imported osm.pbf file. |
values |
logical. |
split |
character. Pattern passed to |
... |
further arguments to |
a list of tags as character vectors, or a nested list of tags and values if values = TRUE.
# See Examples at ?osmclass for full examples # Extract 'other_tags' as list other_tags <- osm_other_tags_list(djibouti_points$other_tags) other_tags[1:10] # Count frequency (showing top 10) sort(table(unlist(other_tags)), decreasing = TRUE)[1:10] # Also include values other_tags_values <- osm_other_tags_list(djibouti_points$other_tags, values = TRUE) other_tags_values[1:10]# See Examples at ?osmclass for full examples # Extract 'other_tags' as list other_tags <- osm_other_tags_list(djibouti_points$other_tags) other_tags[1:10] # Count frequency (showing top 10) sort(table(unlist(other_tags)), decreasing = TRUE)[1:10] # Also include values other_tags_values <- osm_other_tags_list(djibouti_points$other_tags, values = TRUE) other_tags_values[1:10]
Extract Tags as Columns from an OSM PBF Layer
osm_tags_df(data, tags, na.prop = 0)osm_tags_df(data, tags, na.prop = 0)
data |
an imported layer from an OSM PBF file. Usually has a few important tags already expanded as columns, and an 'other_tags' column which compounds less frequent tags as character strings. |
tags |
character. A vector of tags to extract as columns. |
na.prop |
double. Proportion of features having a tag in order to keep the column. |
a data.table with the supplied tags as columns, and the same number of rows as the input frame.
# See Examples at ?osmclass for full examples # Extracting tags of interest (some of which are inside 'other_tags') tags <- c("osm_id", "highway", "man_made", "name", "alt_name", "description", "wikidata", "amenity", "tourism") head(osm_tags_df(djibouti_points, tags)) # Only keeping tags with at least 5\% non-missing head(osm_tags_df(djibouti_points, tags, na.prop = 0.05))# See Examples at ?osmclass for full examples # Extracting tags of interest (some of which are inside 'other_tags') tags <- c("osm_id", "highway", "man_made", "name", "alt_name", "description", "wikidata", "amenity", "tourism") head(osm_tags_df(djibouti_points, tags)) # Only keeping tags with at least 5\% non-missing head(osm_tags_df(djibouti_points, tags, na.prop = 0.05))