Construction monitoring with drones: where control is lost
A weekly photo report from the site foreman is a familiar form of passive control. Photos are taken from the ground, angles are chosen manually, and there is no single point or time of shooting.
Such a set can neither be measured nor compared: the photos show that work is underway, but not exactly how much has been done.
This leads to two systemic failures. Hidden work volumes are not checked against the project — the report says “concrete poured,” but at which work section, in what volume, and with what deviation from the design contours is unknown.
Schedule delays surface not when they can still be caught up, but at acceptance, when deadlines have already been missed.
At the same time, the data is scattered: photos live in messengers and folders, versions do not match, and disputed work is closed either by rework or by payment under a formal certificate. Neither the client nor the contractor can rely on an objective picture of the site.
Drone construction monitoring works differently. The flight follows the same route with coordinate referencing, and the images are stitched into an orthophoto and a three-dimensional model of the site.
The model is superimposed on the design contours, and comparing surveys from different dates gives the volume difference for each work section — what has actually been completed and how much the actual state has diverged from the plan.
Below we break down how this is addressed: how the flight route is built, how often the drone is launched, what goes into the report, and how to use it to make decisions before the delay becomes irreversible.
What regular site flights change: measurable effects
A regular flight is not a one-off photo-reporting exercise, but a data stream that feeds into the project. Each flight produces an orthophoto and a point cloud, from which a digital model of the facility is assembled, tied to the design axes and elevations.
Processing is done in Agisoft Metashape, so comparison with the project becomes geometric rather than “by eye.”
Comparing actual vs. planned using the digital model
The model is superimposed on the design geometry, and discrepancies are calculated by axes, contours, and elevations. The engineer receives a list of deviations with coordinates and volume: what has been installed, where the elevation mark has shifted.
The argument “that’s how it should be in reality” is resolved in one pass — both parties look at the same data set.
Early detection of deviations
Formwork displacement, an error in axis layout, and lag by work sections are visible after one or two flights following their occurrence, rather than at stage acceptance. Rework does not affect adjacent work, and project corrections are made ahead of time.
Work volume and substantiated certificates
Earthworks, concrete, masonry, and roofing are calculated from the model rather than by manual measurement. Certificates rely on measured quantities, so the share of disputed items falls, and differences in interpretation are resolved using the same data.
Fewer management site visits
Dynamics by work section, delays, and deviations are visible in the model and report, so management does not need to be on site every week. Site visits remain for decisions that require presence, not for control for its own sake.
Drones, fixed cameras, or satellite: what to choose for the task
The choice of survey method is determined not by the capabilities of the equipment, but by which question needs to be answered: see the entire site and calculate volumes, capture an event in the moment, or assess the context around the facility. Below is a comparison of three approaches with the honest limitations of each.
| Method | When it suits | Limitations |
|---|---|---|
| Drone flight | A plan of the entire site is needed: earthwork volumes, installation of structures, comparison of actual state with the project. Flight frequency — from once a week to daily during active stages | Depends on weather and flight permits; does not work under canopies or inside buildings |
| Cameras on site | Continuous monitoring of the perimeter, equipment, work regime, and occupational safety — recording is uninterrupted | The angle is set by installation points; structures and equipment block the view; an overall site plan cannot be assembled |
| Satellite imagery | Large territories and hard-to-reach areas where flying is slow and expensive; open Sentinel data provides regular imagery | Cloud cover, updates every few days, lower detail than flights — small elements cannot be distinguished |
In practice, these methods do not compete but complement each other. Cameras capture events and violations in the moment, a weekly flight refines geometry and volumes, and satellite covers external context.
If survey results are brought into a single site model, discrepancies between sources are immediately visible — and it is precisely on them that delay analysis is built.
Implementation stages: from site audit to handover of regulations
We build the work schedule from the state of the site, not from a template: dense development, active cranes, and security regime change both the routes and the permitted flight times.
Below is the sequence of steps after which drone construction monitoring works without our constant presence at the site.
- Site audit. We survey the territory: elevation differences, power lines, restricted zones, takeoff and landing points, and site access regime.
- Coordination of survey windows. We align the flight schedule with crane operation and structure installation so that flights do not interfere with work or create risk for people below.
- Flight routes. We design flight missions with the required image overlap and referencing to ground control points — then images from different weeks align automatically.
- Survey and processing setup. We select the survey mode for the task and assemble an automated pipeline: material upload, orthophoto, point cloud, and stage report.
- Verification on a reference section. We run the full cycle on one work section and compare the result with manual measurement: the discrepancy in volumes and elevations must fall within tolerance.
- Training and handover of regulations. We train your specialists to launch flights and read reports, and hand over working documents: preparation checklists, actions for loss of communication, and material storage procedures.
We review the first cycles together and adjust routes before handover — after that, the system runs on your side.
Handover of materials: model, orthophoto, reports, regulations
Handover is not an archive but a system deployed at your site: the model, images, and reports reside in your environment, are accessible by role, and are updated according to the same regulations we used during the pilot cycles.
Below is the list of what is recorded in the certificate and what can be checked line by line at acceptance.
- Digital model of the facility — georeferenced, assembled from flight cycles: contours, elevations, number of floors, and discrepancies with design documentation for each work section.
- Orthophoto for each flight date — the entire site in a single coordinate system; areas, lengths, and volumes can be measured from it without visiting the site.
- Flight log — date, aircraft, route, image overlap, weather conditions, operator; it shows why a cycle was completed fully or partially.
- Volume report template — comparison of the current cycle with the previous one, deviations by work section, delay zones, and areas requiring re-survey.
- Data access rights — roles are separated: the contractor sees its own work section, the client’s service sees the entire site, and export of source images is restricted by a separate permission.
- Employee instructions — how to accept a flight, how to read the model and orthophoto, and what to do if there is a discrepancy with the completed work certificate.
- Update and storage regulations — flight frequency, history depth by cycles, backup procedure, and deletion of outdated layers.
Acceptance follows these same items: each artifact can be opened, compared with the previous cycle, and handed over to a new contractor or your engineer without our involvement.
Case study: lag across buildings was noticed in the third week of flights
The facility had four buildings at different levels of readiness; flights were done one-off, and the contractor sent a report once a month: certificates, photos from the ground, selective measurements.
Lag across buildings surfaced with a delay of three to four weeks, when rework was already costing schedule. Disputed work volumes were closed by correspondence — each party had its own figures and its own photos.
We moved the facility to a weekly cycle: flight, image processing into a point cloud, and referencing to the design model by axes and floors. The deviation is visible in centimeters and confirmed by images with coordinates, not by wording in a certificate.
We perform surveys using professional DJI aircraft — model comparability from week to week is maintained, so dynamics are read rather than guessed.
Measurements before and after
- Lag across buildings: detected on day 18–25 — now on day 3–4 after the deviation begins.
- Discrepancy between declared and actual volumes: 11–12% — reduced to 2%.
- Resolution of disputed volumes: months of correspondence — now one pass through the model.
The dispute over the third building ended uneventfully: the contractor checked the certificates against the model, removed some volumes, and the schedule was adjusted without stopping work.
From then on, reports were accepted only after comparison with the model — and there were no more reasons for discrepancies.
What to do if your site already has its own surveillance cameras
Existing surveillance cameras at the site are not dismantled — they remain a layer of continuous monitoring: perimeter, equipment entry, warehouses, amenity areas, and stream recording for incident review.
Their limitation is structural: the angle is fixed, and anything that goes into a blind zone behind structures or above the mounting level is not seen by the camera.
That is precisely where data is objectively lacking. A camera looks along the site, while vertical elevations, roofing, facade, work fronts at different sections, and volumes across the entire territory do not fit into its frame.
A single flight provides a slice of the site as a whole and by height; a series of flights provides dynamics between surveys.
Hence the hybrid monitoring scheme: cameras maintain continuity and security, flights provide the state of the facility as a whole.
We separate roles deliberately; otherwise we get data duplication: two sources record the same thing, the volume grows, timestamps and angles diverge, and discrepancies appear in the summary that people stop trusting.
We build the transition to regular flights from what already works. First we mark the zones that are actually covered by cameras and exclude them from routes, then we select survey points and frequency to match the pace of work. Images go into a single chronology, and comparison is from flight to flight, not from scratch.
Cameras continue to record the stream — we are adding a second layer, not replacing the first. Describe what is installed at the site and which tasks are not covered — we will come back with an assessment and a plan.
Technical concerns: communication, weather, flight permits, access
Technical objections to aerial surveying of a site repeat from project to project: permits, weather, communication, data. Below are short answers without promises to “fly in any weather.”
Who prepares the flight permit?
Flights within populated areas and restricted zones require coordination. We handle the documents, route scheme, and interaction with dispatch services — you only need to confirm dates and provide site access.
What to do about weather restrictions?
We fly within the permissible wind and visibility limits for the specific aircraft. In case of a storm warning, we postpone the flight rather than shoot “as it turns out”: a spoiled frame will not align into the model anyway. Missed dates are made up in the nearest window.
Will there be enough communication at the site?
The route is loaded into the aircraft in advance, the flight proceeds autonomously, and the channel is needed for telemetry and commands.
If the signal is lost, the aircraft returns to the takeoff point according to the programmed route — the captured material remains on the storage medium and is not lost in the excavation pit.
How is survey data stored and protected?
Material is uploaded to the client’s or our secure server, access is role-based, and an access log is maintained. Storage periods and export rules are fixed in the contract: data is not transferred to third-party services.
Who is responsible for the aircraft and engineers?
Flights and equipment maintenance are our responsibility, including insurance and pilot certifications. Your side remains responsible for site access and coordinating work with contractors and security services.
Let’s discuss the task: we will select a flight scheme for your site
Development of a monitoring system begins not with code, but with analysis of the site: the flight scheme, survey frequency, and logic of automated processing depend on the geometry of the plot and the stages of work. Therefore, the first step is discussing the task. We need a minimum of input from you; we will clarify and fill in the rest ourselves.
- Boundaries of the plot and key points: what we monitor — earthwork volumes, installation of structures, storage of materials.
- Construction stages and calendar: which works proceed in parallel and where daily detail is needed.
- Required flight frequency and permissible delay between survey and the finished report.
- Regulations for weather windows: what we consider non-flying weather and how we reschedule the flight.
- Reporting forms your services are already used to: logs, certificates, volume tables.
- Site access, approval procedure, and the person responsible for support on flight days.
Then we follow a short cycle: analyze the site and fix the flight scheme for specific tasks, return with an assessment and a stage-by-stage work plan, then launch a pilot flight on one section.
Based on its results, we compare measurement accuracy and report speed with expectations, and only after that do we scale to the entire site. Describe the task — we will come back with an assessment and a plan.
