How Computer Vision Monitors Jobsite Progress From Daily Drone Footage

A drone flying a pre-programmed route over a construction site takes about 20 minutes to capture the entire project. The photos it generates, typically 200 to 500 images per flight, contain more information about site progress than a superintendent could document in a full day of walking the site and taking notes. Computer vision turns those images into structured progress data that updates automatically with each flight.

From Photos to Progress Data

The computer vision pipeline starts with photogrammetry. The overlapping drone images get stitched together into an orthomosaic (a top-down map) and a 3D point cloud of the site. These outputs provide a geometrically accurate representation of everything visible on site at the time of the flight.

The AI then compares this representation against the design model and the previous flight's data. New elements that appear between flights, a section of structural steel, a poured foundation, installed roofing, get identified and classified. The system calculates what percentage of each planned element is now visible and installed.

A national general contractor ran a side-by-side comparison on 12 projects. Each project had weekly drone flights processed through computer vision alongside traditional superintendent-reported progress. The computer vision identified an average of 8 discrepancies per week between reported and actual progress. Most discrepancies were in the 3 to 7 percentage point range on individual activities, small enough to go unnoticed week to week but large enough to accumulate into significant schedule and cost variances over months.

What Computer Vision Can and Cannot See

Aerial computer vision excels at tracking work that is visible from above: foundations, structural framing, roofing, site work, exterior cladding, and any work on open floors before the deck above is placed. On a typical mid-rise commercial project, 60 to 70% of the total construction scope is visible from drone imagery at some point during the construction process.

The technology struggles with interior work after the building is enclosed. Once the roof and exterior walls are in place, a drone flying overhead cannot see the mechanical, electrical, plumbing, drywall, or finish work happening inside. Some projects supplement drone footage with interior 360-degree cameras mounted on robots or carried by workers, but this adds complexity and cost.

Below-grade work presents a timing challenge. Underground utilities, foundations, and waterproofing are visible only during a narrow window before they get backfilled. Frequent drone flights during the early phases of construction are important to capture this work before it disappears underground.

Element-Level Tracking

The most useful computer vision systems do not just measure overall progress. They track individual elements. Each structural column, each section of wall, each area of slab gets its own status: not started, in progress, or complete. This granularity enables precise earned value calculations and accurate schedule updating.

When the system identifies that columns on grid lines A through D are complete but columns on grid lines E through H have not started, the schedule can be updated to reflect the actual sequence of work rather than the assumed sequence. This matters because actual construction rarely follows the planned sequence exactly, and the deviations affect downstream activities in ways that aggregate progress numbers do not reveal.

One project team using element-level tracking on a 300,000 sq ft warehouse discovered that the steel erector was completing bays in a different sequence than planned. The actual sequence was efficient for the erector but was creating access issues for the underground plumbing contractor who needed to work in bays that the erector was skipping. The computer vision data made this coordination problem visible two weeks earlier than it would have surfaced through traditional reporting, giving the team time to adjust the erection sequence before it caused a downstream delay.

Automated Reporting and Dashboards

Raw progress data from computer vision feeds into dashboards that show progress by area, by trade, and by building system. These dashboards update automatically with each drone flight, providing the project team with a current picture of site status without requiring manual data entry or subjective estimates.

The dashboards typically include plan-view overlays showing completed work in color codes, progress curves comparing planned vs. actual, and activity-level status indicators. Project teams using AI construction monitoring tools report that the visual dashboards change the nature of progress meetings. Instead of debating what percentage complete a trade is, the team looks at the actual imagery and data together.

Photo documentation is a secondary benefit. Every flight creates a timestamped visual record of site conditions. This documentation is valuable for dispute resolution, quality verification, and as-built records. Several contractors have noted that the documentation value alone justifies the cost of regular drone flights, with the progress tracking being an additional benefit.

Cost and Logistics

The cost of regular drone flights varies by project size and location. For a project with a full-time drone operator on site, the per-flight cost is essentially zero beyond the operator's salary and equipment costs, typically $60,000 to $80,000 per year fully loaded. For projects using a drone service provider, flights cost $500 to $2,000 per visit depending on site size and complexity.

Weekly flights provide the best balance of data currency and cost for most projects. Daily flights generate more data but the incremental progress between days is often too small for the computer vision to reliably detect. Bi-weekly or monthly flights are sufficient for slower-moving projects like heavy civil work but miss too much activity on fast-track commercial projects.

FAA Part 107 regulations apply to all commercial drone flights, requiring a licensed remote pilot and compliance with airspace restrictions. Projects near airports, heliports, or in controlled airspace need additional coordination. Most established drone service providers handle these regulatory requirements as part of their service.

The Data Integration Challenge

Computer vision progress data is most valuable when it integrates with the project schedule, the cost system, and the BIM model. The integration creates a digital twin that shows not just what the building looks like today, but how today's status compares against the plan and what it means for cost and schedule performance.

Full integration requires that schedule activities map to physical elements in the model, which map to visible elements in the drone imagery. This mapping takes effort to set up initially but pays off throughout the project by enabling automated progress tracking that connects directly to earned value calculations and schedule updates.

The technology is mature enough that early adopter advantages are fading. Drone-based progress monitoring is becoming standard practice on projects over $10 million, and the contractors who are not using it are increasingly at an information disadvantage compared to those who are. The question has shifted from whether to use it to how to use it most effectively.