Sixteen business questions, sixteen filters

0The playground

Stories 1 to 8 use the same file: the BD TOPO 3.5 of Haute-Garonne, exactly as the French mapping agency (IGN) distributes it for free on the Géoplateforme, in a single 2.9 GB GeoPackage with 53 layers. Nothing was clipped beforehand: FilterMate scopes the territory every time.

Source

commune

Selection

Saint-Gaudens (single selection)

Targets

12 layers

Relation

Intersect, no buffer

Result

12,380 buildings, GPKG + ZIP

1. Load the 13 layers and open FilterMate. The Layers panel shows the counts of the whole department: 1,173,829 buildings, 383,112 road segments. Click commune: it becomes the source layer. In the Exploring area, tick Single selection, set the display expression to "nom_officiel" and type "Saint-Gaudens" in the list. The Tracking button of the Exploring column being on, the canvas recentres on the commune by itself.

   

2. Tick every target. Filtering tab, Layers to filter row: right-click in the list, then Select All. The Intersect predicate is ticked, the Buffer row stays off.

   

3. Filter. The log reports under a second for the twelve layers: on GeoPackage, FilterMate first queries the R-tree index of each table, then evaluates ST_Intersects on the candidates only.

   

4. Configure the export. Exporting tab: the 12 layers, EPSG:2154, styles as QML, GPKG format, output folder and a ZIP archive.

   

5. Export. Because the format is GeoPackage and several layers are exported, FilterMate shows a recap and offers to keep the group hierarchy. Confirm: the file is written in 4 seconds with an embedded QGIS project.

   

Result

Under the hood

Provider filter generated for the buildings layer, as read in Layer Properties → Source:

ROWID IN (SELECT id FROM "rtree_batiment_geometrie"
          WHERE minx <= 517706.0 AND maxx >= 511938.3
            AND miny <= 6231762.5 AND maxy >= 6222765.8)
AND ST_Intersects("geometrie", ST_MakeValid(ST_GeomFromText('Polygon ((…))', 2154)))

Source

troncon_de_route

Selection

"importance" IN ('1', '2')

Targets

batiment, erp

Relation

Intersect, buffer by expression, Flat

Result

2,753 buildings, 78 public buildings

1. Scope Muret. Same gesture as in story 1: commune as source, Muret in single selection, every layer as target, Intersect, Filter. Muret has 19,458 buildings, 354 public buildings and 5,512 road segments.

   

2. Pick the main roads. Click troncon_de_route in the Layers panel: it becomes the source. Tick Custom selection and type "importance" IN ('1', '2'). 102 segments are kept out of the commune's 5,512.

   

3. Buffer by expression. Targets batiment and erp, Intersect predicate. On the Buffer value row, open the data-defined property menu, choose Edit…, then enter the following in the expression editor:

   CASE WHEN "importance" = '1' THEN 300
        WHEN "importance" = '2' THEN 250
        ELSE 100 END

   Confirm the editor. On the next row, open the Round / Flat / Square list and choose Flat.

   

   

4. Filter.

   

   

Result

2,753 buildings and 78 public buildings lie in the buffer zones of the 102 segments of importance 1 and 2 of Muret, of which 2,196 buildings and 75 public buildings in the commune itself: the buffer reaches the neighbours, because a new spatial filter replaces the scope (see "One rule to know").

Source

surface_hydrographique

Selection

77 surfaces (Select All)

Targets

buildings, public buildings, roads

Relation

Intersect, 300 then 100 m, Undo

Result

14,426 buildings at 300 m

1. Scope Grenade. 6,998 buildings, 79 public buildings, 1,978 road segments and 77 water surfaces.

   

2. Select every water surface. Click surface_hydrographique, tick Multiple selection, then right-click in the list and Select All. The summary above the list counts 77 features.

   

3. 300 m. Targets batiment, erp and troncon_de_route, Intersect, Buffer value 300. Filter.

   

   

4. 100 m. Change the buffer value and press Filter again.

   

5. Undo. The first button after Filter restores the previous state of the three layers, at 300 m. Redo would reapply it. The history belongs to the project and survives a QGIS restart.

   

Result

Source

troncon_hydrographique

Selection

"persistance" = 'Permanent'

Targets

roads, linear structures

Relation

Cross, no buffer

Result

747 road segments, 364 structures

1. Scope the EPCI. epci layer as source, "Toulouse Métropole" in single selection on "nom_officiel", every other layer as target, Intersect. The filter returns 98,687 road segments, 1,844 watercourse segments, 1,181 linear constructions and 70 touched communes.

   

2. Keep only permanent watercourses. Source troncon_hydrographique, Custom selection "persistance" = 'Permanent': 691 segments.

   

3. Check the selection on the map. Click Identify in the exploration toolbar: the 691 watercourse segments kept by the custom selection flash red. This visual check confirms the filter source before applying it to the roads.

   

4. Cross. Targets troncon_de_route and construction_lineaire, Cross predicate alone, Buffer row off. Filter.

   

   

   

Result

747 road segments and 364 linear constructions (mostly bridges) cross one of the 691 permanent watercourse segments of the metropolis. The same computation done outside FilterMate with GEOS gives the same numbers.

Source

batiment

Selection

"usage_1" = 'Résidentiel'

Targets

zone_de_vegetation, haie

Relation

Intersect, 50 m buffer, then AND

Result

58 woods and forests, 48 hedges

1. Scope Luchon. 3,699 buildings, 269 vegetation zones, 123 hedges.

   

2. Keep only dwellings. Source batiment, Custom selection "usage_1" = 'Résidentiel': 1,619 buildings.

   

3. 50 m. Targets zone_de_vegetation and haie, Intersect, Buffer 50. Filter: 119 vegetation zones and 48 hedges remain.

   

4. Keep only woods and forests. Click zone_de_vegetation in the Layers panel: it becomes the source. Custom selection:

   "nature" IN ('Bois', 'Forêt fermée de feuillus', 'Forêt fermée de conifères',
                'Forêt fermée mixte', 'Forêt ouverte')

   Enable the Combine operator row and leave AND on the left: the new filter narrows the spatial filter already applied to the source layer instead of replacing it.

   

5. Filter.

   

Result

Out of the commune's 269 vegetation zones, 119 touch the 50 m band around a dwelling, among which 58 woods and forests; 48 hedges are in the same case.

Under the hood

The vegetation layer's filter after step 5 keeps the spatial clause and adds the attribute clause:

(ROWID IN (SELECT id FROM "rtree_zone_de_vegetation_geometrie" WHERE …)
 AND ST_Intersects("geometrie", ST_Buffer(ST_MakeValid(ST_GeomFromText('MultiPolygon (…)', 2154)), 50.0)))
AND ("nature" IN ('Bois', 'Forêt fermée de feuillus', 'Forêt fermée de conifères', 'Forêt fermée mixte', 'Forêt ouverte'))

Source

troncon_de_route

Selection

paved roads

Targets

hamlets, activity zones

Relation

Disjoint, 200 m buffer, AND

Result

22 hamlets

1. Scope Aspet spatially for roads, buildings and vegetation (story 1), then by attribute for the two target layers: Toolbox → FilterMate → Filter multiple layers, layers zone_d_habitation and zone_d_activite_ou_d_interet, expression "insee_commune" = '31020'. 87 hamlets and 31 activity zones.

   

2. Keep only paved roads. Source troncon_de_route, Custom selection "nature" NOT IN ('Chemin', 'Sentier', 'Escalier', 'Piste cyclable', 'Route empierrée'): 494 segments out of 1,168.

   

3. Disjoint, 200 m, AND. Targets zone_d_habitation and zone_d_activite_ou_d_interet, Disjoint predicate alone, Buffer 200, Combine operator row on with AND on the right (the target layers' operator).

   

4. Filter.

   

Result

22 hamlets of Aspet are more than 200 m from a paved road; none of the 31 activity zones is. At 100 m they would be 32.

Source

favorite

Selection

settings of story 3

Targets

3 layers

Relation

replayed in one click

Result

14,426 buildings, as in story 3

1. Save. The 300 m filter of story 3 is applied. Click the ★ badge then Add filter, name it "PPRI Garonne 300 m".

   

2. Unfilter. The three layers get the whole department back: 1,173,829 buildings.

   

3. Replay. ★ badge then "PPRI Garonne 300 m": the filter reapplies, with the same 14,426 buildings, 178 public buildings and 6,310 segments.

   

4. Describe and share. Manage favorites… opens the manager: description, tags, scope, exact expression, Remote tab to publish to a git repository. Export… writes a JSON file to send to a colleague, who imports it from the same menu.

   

Source

Processing

Selection

"date_modification" >= '2025-03-15'

Targets

4 department layers

Relation

attribute expression

Result

22,954 modified buildings

1. Open the toolbox and search for "filtermate".

   

2. Choose the layers and the expression. batiment, troncon_de_route, troncon_hydrographique, erp and the expression "date_modification" >= '2025-03-15'.

   

3. Run. The four layers receive the filter in one pass.

   

Result

The same thing from the Python console

import processing
processing.run("filtermate:batch_filter", {
    "INPUT_LAYERS": ["batiment", "troncon_de_route", "troncon_hydrographique", "erp"],
    "EXPRESSION": "\"date_modification\" >= '2025-03-15'",
})

Source

mnt DTM, computed slope (%)

Selection

Slope range ≥ 30 %

Targets

parcelle

Relation

Intersect, coverage ≥ 50 %

Result

58 parcels

1. Compute the slope. Click mnt in the Layers panel, then Compute on → Slope (%). The histogram now shows the slope.

   

2. Pick the range. Tick Value range and type 30 in Min. The share of kept pixels shows under the histogram: 12.9 %.

   

3. Relate to the parcels. Target parcelle, Intersect predicate, cov. ≥ 50 % next to the predicates.

   

4. Filter.

   

Result

58 parcels of the square kilometre have at least half of their area on slopes above 30 %. The same filter on a precomputed slope raster (gdaldem slope -p) gives the same count.

Source

LiDAR HD point cloud (COPC)

Selection

Classification = 6 and height ≥ 9 m

Targets

batiment

Relation

Intersect, coverage ≥ 50 %

Result

57 buildings

1. The building points. Click lhd, tick ATTRIBUTES, + Condition, then Classification = 6. The eye button isolates them on the map.

   

2. The height. Tick Height above DTM ≥, type 9 and pick the mnt DTM.

   

3. Relate to the buildings. Target batiment, Intersect, cov. ≥ 50 %.

   

4. Filter. The height takes 75 s on 20.4 million points, then the filter is instant.

   

Result

57 buildings out of 668 have at least half of their footprint under roof points more than 9 m high. BD TOPO lists 68 at 9 m or more: check the differences on site.

Source

LiDAR HD point cloud

Selection

Classification = 5 and height ≥ 8 m

Targets

troncon_de_route

Relation

Intersect, 3 m buffer

Result

91 segments out of 172

1. The high vegetation. ATTRIBUTES, + Condition, Classification = 5, then the eye to see it.

   

2. Taller than 8 m. Height above DTM ≥ 8 on the DTM.

   

3. 3 m from the road. Target troncon_de_route, Intersect, Buffer 3.

   

4. Filter.

   

Result

91 segments out of 172 have trees taller than 8 m within 3 m: the starting list of the campaign, to rank on site.

Source

batiment

Selection

"nature" = 'Eglise'

Targets

mnt (raster), lhd (point cloud)

Relation

Intersect, 50 m buffer

Result

GeoTIFF 45 KB + LAZ 1.0 MB

1. The church. Source batiment, Custom selection "nature" = 'Eglise'.

   

2. 50 m around, on the DTM and the cloud. Targets mnt and lhd, Intersect, Buffer 50.

   

3. Filter.

   

4. Export. EXPORTING: layers mnt and lhd, folder, ZIP path, Export.

   

Result

A 1.1 MB ZIP: the DTM (GeoTIFF, 45 KB) and the cloud (LAZ, 1.0 MB) of the church and its surroundings.

13Vegetation on an RGB orthophoto

Raster → raster · schematic.

Source

ortho_irc

Selection

NDVI > 0.3

Target

ortho_rvb

Relation

Intersect, no buffer

Expected result

RGB orthophoto masked in a VRT

1. Identify the bands. Choose ortho_irc as the current layer. In EXPLORING, open a colour-role menu and choose IRC if bands 1, 2 and 3 are near infrared, red and green respectively. Adjust the roles if the file uses a different order.
2. Select vegetation. Check RASTER EXPRESSION, then choose Insert → Vegetation (NDVI > 0.3). For a colour infrared image with NIR in band 1 and red in band 2, the expression is:

   ("ortho_irc@1" - "ortho_irc@2") / ("ortho_irc@1" + "ortho_irc@2") > 0.3

   Turn on the preview with the selection button and adjust the threshold by comparing the retained areas with the image.
3. Link to the other image. In FILTERING, check ortho_rvb in Layers to filter, choose Intersect and turn off the buffer. For this first run, start with an unfiltered target and leave combination unchecked.
4. Filter, then export. Hide the source to read the target alone. EXPORTING can write the result as GeoTIFF or COG; Undo restores the complete target orthophoto.

Expected result

An RGB layer clipped to the NDVI mask, with its style and identifier preserved. A threshold of 0.3 is a starting point to check against the season, shadows and image characteristics.

14LiDAR points beneath rooftops

Raster → point cloud · schematic.

Source

ortho_rvb

Selection

Pipette, clicked object

Target

lhd

Relation

Intersect, no buffer

Expected result

Cloud clipped beneath the objects

1. Find the rooftop. Choose ortho_rvb as the current layer and zoom to a roof whose colour stands out from its surroundings. Check the red, green and blue roles in EXPLORING.
2. Detect the object. In the pipette, choose Clicked object, then click inside the roof. Adjust the tolerance while checking the outline on the map. Ctrl+click adds a sample for another part of the roof or another rooftop.
3. Link to the cloud. In FILTERING, check lhd in Layers to filter. Choose Intersect, with no buffer. Start with an unfiltered target cloud to retain every return within the selected footprint.
4. Filter and inspect. Hide the orthophoto to examine the points, then show it again to check their footprint. EXPORTING can deliver LAZ or COPC; Undo restores the complete cloud.

Expected result

A derived cloud limited to the footprint of the retained pixels. This is a spatial filter: every return within that footprint may remain, including ground or vegetation points. To keep only building-class points, then apply Classification = 6 to the resulting cloud.

15The DTM beneath ground points

Point cloud → raster · schematic.

Source

lhd

Selection

Classification = 2

Target

mnt

Relation

Intersect, no buffer

Expected result

DTM restricted to the ground footprint

1. Read the classification. Choose lhd as the current layer. In EXPLORING, open CLASSES and check that class 2 – Ground contains points.
2. Retain the ground. In the Selection menu, choose Ground. Check that only class 2 is selected, then use the selection button to preview the points on the map.
3. Link to the DTM. Check mnt in Layers to filter, choose Intersect and turn off the buffer. Use an unfiltered target and leave combination unchecked.
4. Filter, then read the raster. Hide the cloud to see the clipped DTM. EXPORTING writes the result as GeoTIFF or COG; Undo restores the complete DTM and the previous source state.

Expected result

A DTM VRT whose pixels outside the ground-point footprint are NoData. Existing elevations are preserved: this operation clips the DTM without interpolating it again.

16Vegetation near power lines

Point cloud → point cloud · schematic.

Source

lhd_reseau

Selection

Classes 13 and 14

Target

lhd_vegetation, class 5

Relation

Intersect, 10 m buffer

Expected result

Vegetation within the corridor

1. Prepare two layers. Load the cloud twice in QGIS or duplicate its layer. Name them lhd_reseau and lhd_vegetation. Use a metric CRS for the 10 m buffer, such as Lambert-93 in mainland France. Check that classes 13 – Wire – guard or 14 – Wire – conductor, and class 5 – High vegetation, are present.
2. Prepare the vegetation target. Leave “apply to other tiles in the dataset” disabled on both layers to keep their selections independent. Choose lhd_vegetation as the current layer. In CLASSES, check only class 5. Uncheck Layers to filter and click Filter to apply the selection to this layer alone.
3. Select the wires. Switch to lhd_reseau. In CLASSES, check only the available classes 13 and 14. Use the selection button to inspect the points that will form the source footprint.
4. Clip the corridor. In FILTERING, check lhd_vegetation, choose Intersect and set Buffer to 10 m. Click Filter: clipping preserves the class filter already applied to the target. Hide the source to examine the result, then export the target as LAZ or COPC.

Expected result

A high-vegetation layer restricted to a corridor around the wire-point footprint. The distance is measured in plan view, from a footprint on a grid, to identify areas to inspect. The actual three-dimensional clearance between branches and conductors requires further point cloud analysis.

17Replaying these stories at home

1. Download the BD TOPO of your department from the IGN Géoplateforme ("all themes", GeoPackage format, Lambert-93). The file ships with its spatial indexes: that is what makes the cascades fast.
2. Load the layers named in each story. The styles embedded in the GeoPackage hide buildings beyond 1:25,000: disable scale-dependent visibility in the layer properties, or apply your own styles.
3. Replace the commune names with yours. The fields used (nom_officiel, importance, persistance, usage_1, nature, insee_commune, date_modification) are those of BD TOPO 3.x, identical in every department.
4. To reproduce the screenshots for cases 1 to 8, the repository script scripts/recipe_screenshots.py drives QGIS from end to end of each story, checks the counts and saves one image per step. Recordings for cases 9 to 12 are prepared by scripts/site_videos.py.
5. For illustrated recipes 13 to 16, load the layers named in each recipe and follow the settings. The diagrams show expected results; areas and counts will depend on your data.