Construction progress verification from optical imagery
Very-high-resolution optical satellites can document physical change at construction sites independently of site access or contractor reporting, comparing observed ground conditions against planned build schedules with sub-metre precision.
Sensors
- Maxar WorldView Legion: 30 cm native panchromatic resolution, multispectral at 1.2 m. Up to 15 revisits per day over a target at mid-latitudes when the full six-satellite constellation is operational. Suited to detecting individual structural elements, stockpile boundaries and plant positions.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, four-satellite constellation giving same-day stereo capability. Stereo and tri-stereo tasking enables photogrammetric DSM generation, which is the basis for shadow-independent height estimation.
- Planet SuperDove: 3 m resolution, eight spectral bands (coastal blue through NIR), daily global revisit. Resolution is too coarse for individual structural elements but sufficient for site footprint change, material classification by spectral signature and earthwork-volume estimation on large sites.
- Airbus SPOT 7: 1.5 m panchromatic, 6 m multispectral. Useful for medium-scale infrastructure projects where sub-metre detail is not required and archive depth (imagery back to 2014) matters for baseline comparison.
What a satellite actually sees on a construction site
A very-high-resolution optical image is, at its simplest, a record of surface reflectance at a moment in time. On a construction site that means bare soil, concrete, steel, standing water in excavations, aggregate stockpiles, plant and temporary structures. Each material has a measurable spectral signature. Fresh concrete reflects differently from cured concrete; compacted fill differs from loose spoil. At 30 cm resolution, individual tower cranes, reinforced concrete columns and precast panels are resolvable.
The useful comparison is not a single image but a time series. Differencing two co-registered images from different dates produces a change mask. Pixels that changed significantly in reflectance or texture between acquisitions correspond to physical work done. That change mask can be overlaid on a project schedule to ask a simple question: does the observed footprint of completed structure match the contractor's claimed progress at this milestone date?
Shadow geometry as an independent height gauge
When stereo imagery is unavailable or too expensive to task repeatedly, shadow length offers a practical proxy for structural height. The geometry is straightforward: given the solar elevation angle and azimuth at the moment of acquisition (both recorded in image metadata), the length of a shadow cast by a vertical structure gives its approximate height by simple trigonometry. Published studies using WorldView-2 imagery have demonstrated height retrieval accuracy of roughly one to three metres on isolated structures under clear-sky conditions.
The method has honest limits. Shadows overlap in dense sites, making individual measurements ambiguous. Low solar elevation angles in winter or at high latitudes produce very long shadows that extend beyond the site boundary, complicating measurement. The technique works best on isolated tall elements such as lift cores, chimneys or silos, where the shadow falls on a flat, unobstructed surface. For systematic height mapping across a full site, stereo-derived digital surface models from Pléiades Neo tri-stereo tasking are more reliable, though they cost more and require cloud-free conditions on both acquisition passes.
Material classification and what it can confirm
Multispectral classification assigns each pixel to a surface-material category based on its reflectance across several bands. For construction monitoring the operationally useful classes are typically: bare earth or excavation, concrete (fresh and cured), steel or metal cladding, standing water, vegetation and temporary structures such as hoarding or site offices. Planet SuperDove's eight-band configuration, which includes a red-edge band at roughly 705 nm and a NIR band, gives better separation between construction materials and vegetation than a standard RGB sensor.
Material classification supports progress verification in specific ways. The appearance of a concrete spectral signature inside a previously bare-earth excavation polygon indicates foundation pour. The spread of a metal-reflectance class across a structural frame indicates cladding installation. Neither observation proves quality, but both are independent of contractor reporting. For lenders or project owners monitoring covenant compliance, that independence is the point. The classification is not infallible: spectrally similar materials can be confused, and shadow pixels must be masked before classification or they introduce systematic error.
Revisit frequency versus the pace of construction
Construction moves at different speeds depending on phase. Earthworks on a large site can shift thousands of cubic metres per week; structural steelwork erection on a high-rise may add a floor every five to seven days under favourable conditions. The right sensor depends on which phase the client needs to monitor.
Planet's daily revisit is well-matched to earthworks and bulk material movement, where changes are large and fast but do not require sub-metre detail. WorldView Legion's potential for multiple same-day passes is most valuable for short-duration critical events: a concrete pour, a steel erection sequence or a crane lift that a project schedule flags as a milestone. Pléiades Neo stereo is typically tasked monthly or at milestone dates rather than continuously, because the stereo processing pipeline adds latency and cost. Cloud cover is the unavoidable constraint for all optical sensors. In tropical or monsoon climates, usable imagery may be available only a fraction of the time during wet seasons, and a monitoring programme must account for that by building in SAR-based alternatives or accepting data gaps.
Turning imagery into a progress report
Raw imagery is not a deliverable. The workflow that converts it into a progress report has several steps: geometric correction and co-registration to a common coordinate system, cloud and shadow masking, change detection or classification, comparison against a georeferenced project plan or BIM footprint, and finally a structured output that a project manager or lender can read without specialist training.
The georeferenced project plan is the critical input that clients often underestimate. Without a digital baseline showing where each structure should be at each milestone date, the satellite data can describe what is on the ground but cannot say whether it is on schedule. Satellize structures its analytics workflows to ingest client-supplied schedule data in standard GIS formats and return a milestone-comparison layer alongside the change map. The Overhead column has covered the general methodology for infrastructure monitoring in several public posts. For clients with complex multi-site programmes, the analytics can be scaled across sites using the same processing chain, though each site still requires its own georeferenced plan.
Honest limits of the method
Optical satellite imagery cannot see inside structures, through roofing or below ground. Once a building is enclosed, the exterior envelope is the only observable. Progress on interior fit-out, MEP installation or underground utilities is invisible to any passive optical sensor. For those phases, thermal imaging or SAR-based deformation monitoring addresses different questions, and those methods are covered in sibling pages in this library.
Archive depth is a practical asset that is often overlooked. Maxar's archive extends back to 1999 for some areas; Pléiades archive imagery is available from 2012. For dispute resolution or retrospective schedule verification, historical imagery can establish what was on the ground at a specific past date without requiring any new tasking. The archive is not complete everywhere and cloud cover affects historical coverage just as it affects new acquisitions, but for many sites in temperate or arid climates the record is surprisingly dense.
Typical figures
| Best available spatial resolution | 30 cm panchromatic (WorldView Legion, Pléiades Neo) |
| Multispectral resolution | 1.2 m (WorldView Legion, Pléiades Neo); 3 m (SuperDove) |
| Maximum revisit frequency | Up to 15 passes per day over a target (WorldView Legion, full constellation); daily (SuperDove) |
| Spectral bands | 4–8 multispectral bands depending on sensor; SuperDove adds coastal blue and red-edge |
| Stereo DSM vertical accuracy | Typically 0.5–1 m RMSE for Pléiades Neo tri-stereo under clear-sky conditions (published Airbus figures) |
| Shadow-based height retrieval accuracy | Approximately 1–3 m on isolated structures (published literature range; degrades in dense or shadowed scenes) |
| Minimum detectable change footprint | Approximately 1–4 m² at 30 cm resolution for a high-contrast surface change |
| Cloud constraint | All optical sensors; usable imagery may be <30% of acquisitions in tropical wet seasons |
| Archive depth | Maxar from 1999; Pléiades from 2012; SPOT 7 from 2014 |
| Typical delivery formats | GeoTIFF, shapefile or GeoJSON change layers, PDF milestone report |
Analytics Satellize can run
| Milestone progress map | Bi-temporal change detection (image differencing or classification comparison) against georeferenced project schedule | GeoJSON polygon layer with per-structure progress status; PDF summary report at each milestone date |
| Shadow-based height estimates | Solar geometry retrieval from image metadata; shadow-length measurement; trigonometric height calculation | Point or polygon GIS layer with estimated heights and uncertainty bounds for flagged structural elements |
| Stereo digital surface model | Photogrammetric processing of Pléiades Neo stereo or tri-stereo pairs; dense image matching | GeoTIFF DSM at 0.5–1 m post spacing; volumetric change report comparing successive DSMs |
| Material classification layer | Supervised or semi-supervised multispectral classification (concrete, steel, bare earth, water, vegetation, temporary structures) | Raster classification GeoTIFF; area statistics table per class per acquisition date |
| Earthwork volume estimate | DSM differencing between pre-construction baseline and current surface; volumetric integration | Cut-and-fill volume table; uncertainty estimate based on DSM accuracy |
| Time-series change animation | Co-registered image stack with cloud-masked compositing; change mask overlay | Annotated video or GIF of site progression; tabular change-area statistics per epoch |
| Schedule deviation alert | Automated comparison of classified change footprint against milestone polygon and planned completion date | Email or API alert when observed progress deviates from plan by a client-defined threshold |
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.