Body Screws: Structured Light Catches Loose and Missing
Screws are dense on the front bulkhead, rear bulkhead, chassis and engine bay components of a car body, and a missing, wrong or loose screw weakens the connection. This cell uses a structured light 3D camera that projects a grating pattern and solves phase to build a point cloud, judging screw state on three criteria: height, shape and position. It covers multiple areas within a 25 second cycle, subject to site acceptance results.
Why screws defeat 2D vision
A screw occupies a small share of the component surface, and its contrast against the background is unstable. Oil, dirt and reflection all change the gray distribution, so a 2D image can mistake a stain for a screw or miss a low-contrast screw. This is not cured by tuning, because the criterion depends on brightness, and brightness is changed at will by surface condition.
Height, shape and position answer three different questions
Height answers whether the screw is seated: the height difference between the screw head and the surrounding plane is a hard criterion. Shape answers whether it is the right screw and whether it is fitted correctly, because type and seating attitude leave differences in the profile. Position answers whether it is in the right place; if the mounting position is off, correct height and shape still mean a fault. Combined, the three separate missing, wrong and loose.
What a loose screw looks like in a point cloud
A loose screw sits lower than a fully seated one, and its height distribution is unstable. It is not invisible; it looks shorter and deformed. Height and shape criteria together separate loose from fully seated and fully missing, which a single presence check cannot do.
Four zones, four material behaviours, four parameter sets
Front bulkhead, rear bulkhead, chassis and engine bay differ widely in material and surface state. The cell acquires by zone and adjusts grating parameters to the zone: darker light-absorbing surfaces get higher projection brightness, while light reflective surfaces are adjusted the other way. One parameter set for all zones always leaves one area with unacceptable point-cloud quality.
How 25 seconds covers four areas
Cameras are arranged by area, each area acquires and judges independently, and the results aggregate into a verdict for the body component. Zoning lets each area have its own grating parameters and thresholds instead of compromising every area to accommodate one.
The edge: deep holes and near-identical screws
Structured light depends on grating projection and imaging, so it suits exposed screws that can be projected. Mounting points inside deep holes or fully occluded need camera-angle coverage assessment. Where screw specifications are close and height differences tiny, point-cloud resolution must be checked for enough separation.
Questions before line integration
Why does 2D vision misjudge screws so easily?
Screws are small with unstable contrast, and 2D images are strongly affected by light and surface material, so stains and oil are mistaken for screws while low-contrast screws are missed. A 3D point cloud measures height directly.
How does structured light produce a screw point cloud?
A grating is projected onto the surface and deformed by the screw head. The camera captures the deformed pattern, solves phase to recover height distribution, and builds a point cloud containing the screw form.
What does each criterion judge?
Height judges seating, shape judges type and assembly attitude, position judges whether the point is correct. Together they distinguish missing, wrong and loose.
Can a loose screw be identified?
Yes. It sits lower than a seated screw with an unstable height distribution, which height and shape criteria can separate. Detection capability is subject to site acceptance results.
Does this suit every body area?
It suits exposed, projectable screws. Deep-hole or fully occluded mounting points need coverage assessment by camera angle and cannot be assumed.