Building permit compliance verification against approved footprints
Very-high-resolution satellite imagery lets planners and valuers compare what was built against what was permitted, catching unauthorised extensions, extra storeys and boundary encroachments before they become legal liabilities.
Sensors
- Maxar WorldView-3: Panchromatic resolution of 0.31 m, multispectral at 1.24 m. The sharpest commercially available spaceborne optical sensor at scale; capable of delineating wall lines and small rooftop plant to within roughly one metre on the ground.
- Airbus Pléiades Neo: 0.30 m panchromatic, 1.2 m multispectral, four-satellite constellation offering same-day or next-day revisit over most cities. Stereo and tri-stereo collection modes support 3-D height estimation, which matters when the violation is a new storey rather than a footprint extension.
- Planet SkySat: 0.50 m panchromatic, 1.0 m multispectral. Slightly coarser than WorldView-3 or Pléiades Neo but with a larger constellation and flexible tasking windows. Adequate for footprint comparison on buildings wider than about three metres; less reliable for narrow-plot terraced housing.
- Airbus SPOT 6/7: 1.5 m panchromatic, 6 m multispectral. Insufficient for individual wall-line delineation on dense urban plots, but useful for screening large areas to prioritise which parcels warrant a higher-resolution follow-up tasking.
What the approved drawings do not show you
A building permit describes an envelope: a footprint on a site plan, a ridge height, a setback from the boundary. What gets built is a physical object that can be measured from above. The gap between those two things is where enforcement agencies and mortgage valuers spend a surprising amount of time.
The problem is not rare. Local authorities in England alone recorded over 10,000 enforcement notices in a single recent year, the majority relating to works carried out without permission or outside consented parameters. Mortgage lenders face a related exposure: a property sold with an unapproved rear extension has a title defect that can affect resale and insurance. Neither the council nor the lender has historically had a fast, systematic way to check at scale.
Resolution is not optional here
The physics of this problem are unforgiving. A typical residential rear extension might project two to three metres beyond the original rear wall. To detect that reliably from space, you need imagery where a single pixel represents no more than about 0.5 m on the ground, and ideally 0.3 m. That is exactly the specification of WorldView-3 and Pléiades Neo. Landsat and Sentinel-2, at 10 to 30 m per pixel, are simply the wrong tools for this job.
Even at 0.3 m resolution, automated footprint extraction has limits. Dense vegetation overhanging a rear extension can obscure the true wall line. Flat-roof additions at the same height as the original structure are harder to detect from a nadir (straight-down) view than from a stereo pair that yields a 3-D surface model. Pitched-roof dormers are detectable in plan if they change the ridge line, but subtle internal loft conversions that do not alter the external envelope are invisible to any spaceborne sensor.
The honest summary: satellite imagery reliably catches footprint expansions, boundary encroachments and additional storeys that alter the roofline. It does not catch internal subdivision, change of use or works hidden beneath an unchanged roof.
Matching as-built to as-permitted
The core workflow has three steps. First, extract a building footprint from the satellite image using either manual digitisation or a trained segmentation model. Convolutional neural network approaches trained on labelled very-high-resolution imagery can achieve footprint delineation accuracy of roughly one to two metres under good conditions, consistent with published benchmarks from the ISPRS 2D Semantic Labelling benchmark and similar open evaluations. Second, register the extracted footprint to the coordinate reference frame of the approved permit drawing, which is typically supplied as a PDF site plan or a CAD file. Georeferencing that drawing to sub-metre accuracy requires at least three reliable ground control points visible in both the drawing and the image. Third, compute the geometric difference between the two polygons and flag parcels where the as-built footprint exceeds the permitted envelope by more than a chosen tolerance.
Tolerance matters. A one-metre positional uncertainty in the imagery, combined with a similar uncertainty in the permit drawing's georeferencing, means that differences smaller than about two metres should be treated as ambiguous rather than as confirmed violations. Enforcement agencies need to set their own threshold for what triggers a site visit versus what gets filed as within acceptable measurement error. Satellite analysis generates a prioritised list; it does not replace the inspector.
Height violations need a third dimension
Footprint compliance is the easier half of the problem. Storey additions are harder because a plan-view image shows you the roof, not how many floors sit beneath it. Two approaches exist in the public literature. Stereo photogrammetry from Pléiades Neo or WorldView-3 stereo pairs produces a digital surface model with vertical accuracy of roughly one to two metres under good conditions. That is sufficient to detect an added storey (typically 2.5 to 3 m of additional height) but not to count floors precisely on a building where each storey is close to that uncertainty range.
Shadow-length analysis offers a complementary method for nadir-only imagery: the length of a building's shadow at a known solar elevation angle encodes the object's height. The method is well-documented in the remote sensing literature and works reasonably well on isolated structures with clear shadow falls. It degrades in dense urban canyons where shadows overlap, and it requires precise knowledge of the acquisition geometry, which reputable imagery providers supply in the metadata.
Practical limits and what to do about them
Cloud cover is the operational constraint that does not go away. Pléiades Neo's four-satellite constellation improves revisit, but a persistent overcast over a city can delay usable acquisition by days or weeks. For enforcement programmes with statutory deadlines, this is a genuine scheduling risk that needs to be planned for.
Archive depth is an asset. WorldView-3 has been collecting since 2014, Pléiades since 2012. For a disputed extension where the owner claims it pre-dates the planning system's jurisdiction, a time series of archived imagery can establish when the structure first appeared. That is useful evidence, though its admissibility in legal proceedings depends on jurisdiction and how the chain of custody is documented.
Satellize runs permit-compliance screening as a structured analytics product, georeferencing permit drawings, running footprint extraction and delivering a flagged GIS layer with per-parcel difference polygons. The workflow is similar in principle to the spatial change-detection methods used in the Tonga crop-estimation programme, adapted for hard urban boundaries rather than agricultural field edges.
Who uses this and what they actually receive
Municipal planning enforcement teams are the primary users. A typical output is a ranked list of parcels sorted by the area of the detected exceedance, with a per-parcel report showing the approved footprint polygon, the extracted as-built polygon and the difference geometry highlighted. Teams can load this directly into their GIS and cross-reference against their enforcement case management system.
Mortgage valuers and conveyancing solicitors use a lighter version: a binary flag indicating whether the satellite-derived footprint is materially consistent with the approved drawings, plus an uncertainty band. That flag does not replace a surveyor's inspection, but it can triage a large portfolio quickly. For a lender reviewing a book of properties before a securitisation, the cost of systematic satellite screening is small relative to the cost of a single title defect discovered post-completion.
The next concrete step for a planning authority is to share a sample of permit drawings from a defined area, agree a positional tolerance threshold, and run a calibration pass against a set of parcels with known outcomes. That calibration determines the false-positive rate before any enforcement action is taken.
Typical figures
| Best available spatial resolution | 0.30 m panchromatic (Pléiades Neo, WorldView-3) |
| Footprint delineation accuracy | Typically 1–2 m under good conditions; degrades with vegetation overhang or shadow |
| Minimum detectable footprint exceedance | Approximately 2 m beyond permitted envelope (below this, measurement uncertainty dominates) |
| Vertical accuracy (stereo DSM) | 1–2 m RMSE from Pléiades Neo or WorldView-3 stereo pairs; sufficient to detect added storeys of ≥2.5 m |
| Revisit (commercial tasking) | Same-day or next-day over most cities with Pléiades Neo four-satellite constellation |
| Archive depth | WorldView-3 from 2014; Pléiades from 2012; SPOT 6/7 from 2012 |
| Spectral bands used | Panchromatic (footprint extraction); RGB and near-infrared (vegetation masking, material classification) |
| Cloud cover constraint | Optical sensors only; persistent overcast can delay acquisition by days to weeks |
| Delivery formats | GeoPackage or Shapefile (footprint polygons), GeoTIFF (difference raster), PDF per-parcel report |
| Coverage per tasking pass | Up to ~2,500 km² per strip for Pléiades Neo; practical urban screening areas typically 50–500 km² |
Analytics Satellize can run
| As-built footprint polygon layer | CNN-based semantic segmentation of VHR imagery (e.g. U-Net class of architectures, consistent with ISPRS benchmark approaches) | GIS polygon layer, one feature per building, with area and perimeter attributes |
| Permit exceedance difference polygons | Geometric overlay of as-built and georeferenced approved-permit polygons; difference computed in local projected coordinate system | Flagged GIS layer with exceedance area (m²) and linear overhang (m) per parcel |
| Ranked enforcement priority list | Threshold-based ranking by exceedance area, filtered by positional uncertainty band | Spreadsheet or database export sorted by exceedance magnitude, ready for case management import |
| Storey-addition detection via stereo DSM | Stereo photogrammetry from Pléiades Neo or WorldView-3 stereo pairs; height difference against baseline DSM or permitted ridge height | Per-parcel height-change report with before/after DSM sections |
| Shadow-length height estimate | Solar geometry calculation from image acquisition metadata; shadow length measured in image; height derived trigonometrically | Estimated building height attribute appended to footprint layer, with uncertainty range |
| Temporal first-appearance analysis | Change detection across archive image time series to establish date range when structure first became visible | Per-parcel timeline report with annotated image thumbnails showing before/after state |
| Boundary encroachment flag | Spatial intersection of as-built footprint with cadastral parcel boundary; buffer analysis against consented setback distances | Binary flag plus measured encroachment distance (m) for each affected parcel |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.