Ortholens, from BlackBox: 39N, automatically detects and tracks tension cracks in the drone photogrammetry imagery you capture — on whatever cadence fits your program. Findings go to your PE of record, not a dashboard for the public.
You're already running the Annual Inspection under §257.83 and the five-year Periodic Structural Stability Assessment under §257.73(d). Both are snapshots — a qualified professional walks the embankment, documents conditions, and signs off.
A tension crack can initiate and propagate entirely inside that interval. The gap isn't a knowledge problem — it's a sampling-frequency problem on a slope whose failure mode doesn't run on your inspection calendar.
EPA CCR RULE · 40 CFR PART 257 →Every client's imagery, findings, and models are siloed. The models that analyze your facilities are trained on your company's data only, and nothing learned at your site is ever carried over to another client. A structural finding at your facility goes to you and your PE of record — full stop.
Your imagery, findings, and reports are kept in an environment dedicated to your company — never commingled with another client's data or shared with any third party.
Detection models for your facilities are trained exclusively on your company's data. No other client's imagery goes into your model, and yours never goes into theirs.
Nothing learned at your site — imagery, labels, or model weights — is ever transferred to another client's model. What we find at your facility is disclosed to you and your engineers only.
Your Annual Inspection and 5-year Periodic Assessment stay exactly as they are. This fills the interval between them — it isn't a substitute inspection.
Findings go to the engineer responsible for your facility — never acted on autonomously, never surfaced to anyone outside your team.
Reports export in a form suitable for your CCR compliance file and available for state or EPA review, alongside your existing inspection records.
Deployed at the pond or unit level using imagery you already have access to or can readily acquire — no new instrumentation required to begin.
We don't fly drones or capture imagery — you do, on whatever schedule fits your program. Our image-segmentation algorithm automatically analyzes every drone photogrammetry survey you send us, detecting and tracking tension cracks as an early-warning signal for slope failure and erosion. It's a supplement to your inspection program, not a replacement for the PE judgment it's built to support.
Every survey you capture is analyzed automatically — no manual review of raw imagery required, on the cadence you choose.
Tension cracks are identified and tracked before they progress to visible failure — when there is still time to act.
Every finding is advisory — reviewed by qualified engineers who determine the response. Never autonomous action.
A single image tells you a crack exists. A well-trained model tells you its size. A history of images tells you where it's headed. All illustrated schematically below.
The model identifies the presence and rough location of a candidate feature — enough to tell you where to look.
Higher-fidelity imagery and a more capable model resolve the full extent of the crack — precise length, width, and orientation.
Tracked across surveys, growth rate becomes a pattern — projected forward, toward an estimated failure threshold.
The name describes the two halves of the product: the imagery processing that turns a drone survey into a measurable map, and the machine-learning analysis that finds what matters on it.
Each survey is stitched into an orthomosaic — a single, geometrically corrected map of the embankment, georeferenced so that a finding has a real position on the ground and can be measured and compared from one survey to the next.
A model trained to see one specific signature on that map. Today's lens detects tension cracking in coal ash embankments. The same pipeline can be trained, on your data, to find other signatures your program needs to track.
Have a feature you need to find across your surveys? We'll scope whether a lens can be trained for it.
Request a consultation →You (or your contracted UAV provider) capture drone photogrammetry imagery on the cadence that fits your program. We don't fly drones or collect imagery ourselves.
Each survey is automatically segmented for crack formation and progression indicators, processed for your report only.
Flagged features are delivered to you, confidentially, in a structured report with confidence scoring.
Qualified personnel review the findings and determine the appropriate response.
Built for engineers, not instead of them.
Every output is advisory. Every finding requires qualified human review. BBX augments the engineering judgment it helps apply — it does not act on its own.
Our approach builds on a concept documented in peer-reviewed, publicly available research. BlackBox: 39N's implementation — the models, the training pipeline, and the reporting pipeline — is our own, purpose-built for this surface type, this crack morphology, and this failure mode.
This is the shape of what you'll receive — status, findings, and an engineer's sign-off. The content is deliberately blocked out here: your report will contain your data, reviewed by your team, seen by no one else.
We work directly with utility engineering teams, asset integrity groups, and the engineering firms supporting them — not procurement generalists. If there's a signature beyond tension cracking you need to track, ask about a consultation. Available under NDA before any imagery changes hands.