Transparent parts and 3D vision: how to run a feasibility trial

Transparent 3D scanning for robot guidance needs a feasibility trial before a production commitment. A clear bottle, protective film, and translucent molded component can present very different imaging problems. Begin with the feature the robot needs to locate, not a demand to reconstruct every visible surface.

Define the required feature

List the decision the image must support: identify a container, locate its rim, estimate a pickup surface, confirm a component is present, or measure a dimension. These tasks need different evidence. A useful outline may support a guided pick even when a full surface model is incomplete; dimensional inspection may require a different setup.

Zivid transparent-object imaging guidance explains that transmitted light can create missing or background-associated depth data. Treat attractive point-cloud screenshots as a starting point. Ask the supplier to show whether the actual contact surface is located correctly in the relevant directions, including conditions in which the algorithm should decline a pick.

Read the 2D and 3D selection guide before specifying a depth camera. If a stable fixture makes height known, a silhouette or other visible feature may solve the task with less complexity. That possibility should be evaluated, not assumed.

Assemble a sample matrix

Send a deliberately varied sample set. Include allowed wall thicknesses, colors, labels, seams, closures, surface wear, and relevant liquid contents. If condensation or protective film appears in normal production, include it in the scope. Identify contamination that is unacceptable so the trial does not accidentally normalize a defective process.

Keep every image paired with a sample identifier and a known setup. Without that record, a vendor can improve the image while unintentionally excluding the condition that originally caused the problem. Use repeatable fixtures for the trial even when the eventual production presentation will vary.

Compare optical options without assuming a cure

Edmund Optics light polarization guide describes polarization as a way to manage certain bright reflections. Ask whether it improves your required feature, how it affects exposure, and whether its benefit persists across product orientations. Do not assume a polarizer makes every transparent object suitable for every 3D technology.

Run a small controlled experiment: change one optical or mechanical variable at a time and preserve a baseline configuration. Record usable feature coverage, location error at checked points, acquisition time, and false targets. The reflection-control guide provides a similar structure for separating lighting effects from software tuning.

KEYENCE displacement sensor selection distinguishes confocal and triangulation measurement approaches. For a narrow dimensional question, ask a measurement specialist whether a dedicated sensor is a better fit than a general scene camera. The fact that a device produces 3D data does not establish that its output is adequate for your inspection tolerance.

Verify the physical result

Hypothetical example: a cell must pick clear lids from a tray. A test produces a recognizable tray and lid outline but inconsistent lid-top height. Rather than declaring the image successful, compare the estimated pickup plane with independently checked positions and run controlled grip trials. A fixture change, alternate pickup feature, or different sensor may be the appropriate next step.

Set acceptance around the operation: correct part selected, intended surface reached, grip established, and part delivered undamaged. Count wrong-object picks and confidently incorrect locations separately from explicit “no result” responses. A system that honestly requests assistance can be easier to manage than one that produces plausible wrong answers.

Write the purchase boundary

Approve only the tested material and presentation envelope. Document excluded conditions, a method for adding new products, and who owns future qualification. Request raw test data, settings, and the chosen recovery behavior as handover deliverables.

Use the 3D camera selection checklist to compare quotes after the trial. The objective is reliable handling of your defined products, not an unqualified promise that the camera can see through anything.

If you are evaluating FAIRINO to handle clear components, make the optical trial an early project decision. Use the FAIRINO selection guide to define handling load and reach while the vision supplier tests the required gripping feature. Fixture-based presentation may also deserve a quote if it removes a difficult sensing requirement.

Request a FAIRINO handling quote with clear-part samples, approved surface variations, and the proposed pickup method. Include the imaging trial’s pass conditions so the robot and sensing proposal address the same production task.