What a 360° recording is
A 360° camera has two or more lenses facing opposite ways, each covering more than a hemisphere, and it stitches their images into a single sphere. Played back, the viewer can look in any direction from the camera's position, as if standing there. Recorded as video while the camera moves along the blade, the result is a continuous sphere at every position along the drive: every surface, from every angle the camera passed, whether or not anyone was looking at it at the time.
Detectra records with an Insta360 X5 in 8K, 7680 × 3840 pixels, at 24 frames per second, stitched on board and written to 250 GB of local storage. One pass of the crawler records the whole inside of the blade along its route.
What 8K means on the laminate
Resolution across a sphere is different from resolution in a normal photograph. The 7680 pixels of an 8K frame are spread around 360 degrees, so there are about 21 pixels per degree of view. How many millimetres one pixel covers depends on how far the surface is from the camera. At one metre, a degree of view is about 17.5 mm wide, so a pixel covers roughly 0.8 mm. At half a metre, roughly 0.4 mm.
Inside a blade the camera is rarely more than a metre from the web or the shell, and often much closer. So a 360° recording resolves features below a millimetre on the surfaces that matter, before stitching losses, without anyone aiming at them. A hairline crack in an adhesive fillet is within reach of the resolution. Whether it is visible depends on the light.
Why the light matters more than the lens
Every surface inside a blade is glossy resin, and the camera is the only light source. A crack, a wrinkle or a step at a debonded web foot is a change in the shape of the surface, not a change in its colour. Lit from the camera's position, the surface reflects evenly and the shape is invisible. Lit from the side, the shape casts a shadow and stands out.
That is why Detectra's four LED panels, 52 LEDs each, are switched individually rather than all on at once. The operator lights the surface from one side, records, and where needed lights it from the other side and records again. The recording carries both, and the reviewer chooses. It is also why overexposed footage is worse than dark footage: a specular highlight on glossy resin destroys the detail under it.
What the distance readout adds
A frame of video with no position is a photograph. Every Detectra take carries a telemetry overlay with the distance travelled from the root, and every recording carries its metadata: site, turbine, blade, chamber, operator and time. That does three things:
- A finding has a location. "Web-to-shell bond, pressure side, 23 m from the root" is a place a repair technician can go to.
- The next inspection can return to it. The same distance on the next recording shows the same place, so growth or stability can be seen side by side.
- The fleet can be compared. The same finding at the same distance on several blades of one type is a pattern, and patterns are what serial-defect and warranty conversations are built on. See End-of-warranty blade inspections.
Reviewing at the desk
The recording is uploaded to Fenestra after the visit. The reviewer scrubs along the blade in the 360° viewer, looks in any direction, and flags a finding on the frame where it is seen; the distance readout is attached to the flag. AI proposes candidates along the way, and the reviewer confirms, rejects or reclassifies each one before it can enter a report. Review takes about 3 hours per turbine today.
Three habits make review better:
- Look at the bond lines twice, once with each lighting direction, because a crack that is invisible in one is obvious in the other.
- Compare against the previous recording of the same blade at the same distance before deciding whether a finding is new.
- Flag what is not a defect, such as adhesive squeeze-out or a manufacturing marking, so the next reviewer does not rediscover it.
What a recording cannot do
A camera sees surfaces. Depth, a delamination below the surface or the true extent of a crack, needs a measurement method, which is what the second generation of crawler, Quantra, is being built for. NDT for wind turbine blades covers the methods. And a 360° recording of a whole blade is a large file: 250 GB of storage on the crawler is there for a reason, and the upload is done from the customer's PC after the visit, not from the turbine.
Frequently asked questions
Why use a 360° camera instead of a normal camera inside a blade?
Because nothing has to be aimed at while the camera is in the blade. The full sphere around the camera is recorded at every position along the drive, so the looking is done afterwards, from any angle, at a desk, and nothing is missed because someone did not point at it.
How much detail does an 8K 360° recording show?
About 21 pixels per degree of view. At one metre from the surface a pixel covers roughly 0.8 mm; at half a metre, roughly 0.4 mm. Inside a blade the surfaces are usually within a metre, so sub-millimetre features are within the resolution, before stitching losses.
Why does lighting matter so much inside a blade?
Every surface is glossy resin and the camera is the only light. Cracks, wrinkles and steps are changes in shape, not colour, and only show when lit from the side so that they cast a shadow. Light from the camera's position flattens them and the reflection hides them.