Calibration boards for robot vision: what to buy and validate

When buying a vision-guided robot cell, specify its calibration board against the placement task and calibration procedure. The right target must work with your software, remain stable in your environment, and occupy a useful part of the camera view. A board that produces a successful software message can still leave your robot missing a fixture.

Define which calibration you are buying it for

Ask the integrator to name the result the board will establish: camera intrinsics, a camera-to-workplane mapping, or the relationship between camera and robot. These are different deliverables. MVTec 3D and hand-eye calibration describes hand-eye calibration as the camera-to-robot relationship. A quotation that says only “camera calibrated” leaves the robot placement requirement unresolved.

Document the intended working distance, closest and farthest objects, operating field of view, and required placement tolerance. Include the actual lens and focus setting. For a moving camera, identify the imaging poses and the tool configuration. Our camera-mounting comparison helps separate fixed-camera requirements from wrist-mounted arrangements.

Write a target specification

OpenCV camera calibration tutorial supports multiple target patterns and distinguishes squares from detected inner corners. Confirm exactly which count the software expects before ordering. A dimensional misunderstanding is easier to prevent in the purchase order than to diagnose after a camera has been installed above a machine.

Do not choose the largest board automatically. Request a proposed capture plan showing that the target remains usable at the required distances and angles. Check physical access: an excellent laboratory target is inconvenient if technicians must remove guarding or dismantle the gripper to position it during authorized maintenance.

Decide whether a printed target is sufficient

A printed target may be appropriate for an early feasibility exercise. For production purchasing, ask whether printer scaling, mounting flatness, moisture, wear, and cleaning can consume a meaningful share of your error budget. Require the supplier to explain how the target geometry is verified. Avoid treating a high-resolution image file as a certificate for the physical object that came out of a printer.

Keep the prototype and production decisions separate. It is reasonable to prove that the camera can see the feature before purchasing a durable reference, provided the final acceptance test uses the production setup. Include replacement availability and a storage case in the quote if the target will move between workstations.

Verify a result outside the calibration data

Cognex In-Sight calibration guidance calls for consistent geometry between calibration and operation. Accordingly, record lens settings, mounting position, and target identity in your change-control sheet. Make a moved bracket or replaced lens a reason to review the validation procedure, rather than leaving recalibration to someone’s memory.

For acceptance, propose independently checked locations at the center and edges of the usable workspace and at relevant heights. Compare the robot’s resulting placement with the process requirement. Record both repeated trials at one point and trials spread across the workspace. The distinction matters; see repeatability versus accuracy before accepting one small number as proof of overall performance.

Hypothetical example: a tray cell places parts acceptably in its center pockets but misses pockets on one side. Repeating the center test 100 times would not answer whether the mapping is adequate. A grid of checked pockets, with the operating tool and representative parts, creates a more useful buying decision. This is a proposed test design, not a predicted result for any camera.

Specify the handover package

• The target drawing, identifier, handling instructions, and replacement source.
• Calibration images, software version, settings, and coordinate conventions.
• Independent validation results with the agreed acceptance limits.
• Instructions for verification after maintenance and responsibility for approving changes.

Place these deliverables in the machine vision integration plan. The goal is a procedure your team can repeat, not a board that looks precise in a cabinet.

For a planned FAIRINO cell, include calibration and independent placement checks in the robot specification. Use the FAIRINO selection guide to define the arm’s task, then identify who will verify the complete camera, tool, and fixture relationship. A board purchase alone does not establish cell accuracy.

To request a FAIRINO robot quote, provide the workspace dimensions, placement tolerance, camera-mounting concept, and maintenance constraints. Ask for calibration responsibilities and acceptance evidence to be identified alongside the proposed robot.