Industrial 3D cameras: a practical selection checklist

An industrial 3D camera for a robot cell should be selected against the working volume and handling task you intend to buy. The camera must see the relevant features where they actually occur, at a speed and reliability the cell can use. Begin with samples and layout drawings; leave model selection until the operating requirements are clear.

Draw the usable volume

Mark the closest and farthest required surfaces, not just the center of the workspace. Include full and nearly empty containers, maximum stack height, and the edges that matter to gripping or measurement. Add nearby robot links, tool geometry, guards, and machine doors to the layout.

Zivid bin-picking occlusion guide explains how objects and bin walls create hidden regions. Ask the supplier to show usable coverage at the difficult locations. A field-of-view rectangle does not prove that every important surface inside it will produce valid depth data.

For a wrist-mounted camera, include all intended capture poses and the mass of its bracket and cables. For a fixed camera, review maintenance access and whether normal replenishment can move the mount. Keep the camera’s observation needs separate from the robot’s reach envelope until both have been checked together.

Choose a method with samples

IDS projected texture stereo principles describes projected texture as assistance for stereo correspondence. That is one sensing approach, not a universal solution. Ask each supplier why its selected method fits your surface finish, product geometry, motion, and ambient environment. Compare actual results rather than a diagram of the principle.

Use a sample matrix spanning allowed materials and finishes. Include black, glossy, textured, or translucent items if they are part of the job. Record where data is absent and where it is confidently wrong. Our transparent-part feasibility guide gives additional conditions for clear products; our stereo camera guide examines correspondence and mounting tradeoffs.

Separate specifications that answer different questions

Ask what every quoted number describes: pixel count, spatial sampling, repeatability, accuracy, or maximum acquisition rate. Require the conditions used to obtain it. A single favorable figure should not substitute for the result at the farthest working distance on your most difficult allowed product.

EMVA 1288 camera characterization standard standardizes characterization of image sensors and cameras. Such documentation is useful when comparing imaging components; it does not by itself validate a complete robot-guidance or dimensional-measurement system. Keep application acceptance in the purchase agreement even when the component specifications are well documented.

Request raw data as well as processed output. Filtering can make a surface look cleaner, so you need to know whether it also removes a small feature important to gripping. Have the supplier identify settings that may affect geometry or exclude low-confidence regions.

Verify throughput in context

Hypothetical example: a camera captures quickly, but the robot must move to a clear observation pose and wait for the scene to settle. A nominally slower fixed-camera arrangement might complete the overall task sooner. Measure the full sequence and avoid assuming that frame rate equals production rate.

Include data transfer, processing, result validation, and any repeated acquisition. If multiple cameras share the cell, have the supplier evaluate their interaction and network load. Check behavior during lost frames and communication interruptions through the designed fault-recovery procedure.

Compare the complete package

List the camera, optics where separate, power, cables, mounting, computer, licenses, protective hardware, calibration tools, integration, and training. Ask what changes when you add a new product or replace a camera. The 3D camera cost guide turns those items into comparable quotes.

Choose the setup that proves the task across the agreed operating envelope and has an owner for future support. A well-documented limitation is more useful than an unsupported claim that every surface will work.

For a FAIRINO cell, review the camera’s working volume together with the robot layout. A candidate arm in the FAIRINO catalog should be evaluated with the gripper, imaging pose, and transfer path drawn in place. Keep visibility checks distinct from robot reach checks: both must pass for the same supported parts and container positions.

Request a FAIRINO vision-guided handling quote with the cell drawing, camera trial results, and mounting preference. Include any remaining blind spots or untested surfaces so the proposal addresses them explicitly.