3D vision for robot bin picking

Can a robot reliably pick and place your part?

Share the part, how it arrives and where it must go. We’ll outline which vision method can guide the robot and what to check for the pickup, motion and placement.

Robot gripper taking a machined part from a bin
Illustration · Robot lifting a machined part from a bin

Why MentiVision

Engineering decision support for industrial vision

Supplier leads screened
805
Supplier sites reviewed in depth
42
Hardware models indexed
133

We help engineering and procurement teams decide which approach fits a part, a tolerance and a production line. We compare the options, explain what would rule each one out, and identify which test could change the decision.

Our research on Chinese industrial vision suppliers gives the recommendation a concrete basis: model specifications, application examples and documented limitations. We flag what is known, what depends on your samples and what a supplier still needs to confirm.

What an application review gives you

  • A route to evaluate first
  • The trade-offs and evidence behind it
  • A test that can change the decision

The handling example compares the published working ranges and fields of view of the Mech-Eye PRO S-GL and TransferTech Laser L V2S, then identifies the bin coverage, gripping and placement tests to run.

See an example review →See how the robot cell fits together →
What these figures cover

November 2026 research: 805 deduplicated leads screened against the application scope, 42 candidate sites reviewed in more depth and 133 hardware model records indexed from manufacturer materials. The index includes 106 area-scan models from one manufacturer.

We also mapped nine public support, testing and integration leads in the US and Japan. For a live project, we check which provider can take the work and confirm scope, availability and responsibility before any handoff.

What you get from a review →How we can help →See a real model comparison →

What you gain before committing to equipment

Get a route shortlist, the evidence behind it and the next test from one project brief.

Screen in less engineering time

Compare published specifications and rule out poor fits before starting separate supplier conversations.

See the routes worth comparing

The deep-bin example compares two camera options against working distance and coverage, then identifies where the gripper and fixture need separate tests.

See the costs that change the decision

Compare equipment, fixtures, software, integration and support on the same scope. Spot scope differences before you decide which quotes to pursue.

Reuse one project brief

Share the part, acceptance rule and line constraints once. Use the same summary to brief your team, request quotes and plan a sample test.

See how the review applies to a project →

Work out what the robot cell needs to prove

Separate the jobs inside the cell

Identify where the project needs depth imaging, pose correction, gripper design or robot integration.

Compare published specifications in context

Use working distance, field of view and installation geometry to screen 3D cameras from Chinese manufacturers.

Plan a test with meaningful outcomes

Check the full cell against actual parts, placement tolerances and sustained cycle time.

Compare the selection criteria ↓

Explore three vision-guided handling tasks

Compare how each task works, what the manufacturer materials support, and what to verify with real parts.

Choose by what vision must resolve. If vision must correct the pose at the destination, choose alignment. Otherwise, choose bin picking when vision must locate a pickable part in loose supply, or transfer when a tray or fixture already presents the part.

  1. Capture depth
  2. Plan a grasp
  3. Place or retry
Pick from loose supply
  1. Capture depth

    Fixed 3D camera above the bin

    Scan the pile from above.

  2. Find a pickup pose

    Illustrative point cloud with a candidate bracket pose

    Identify an exposed part and estimate its pose.

  3. Pick the part

    Robot arm lifting a selected metal part from the bin

    Plan a clear approach and lift one part.

Illustrative sequence · The vision output guides the robot action.
How it works
The 3D camera measures depth across the pile. The vision system uses the point cloud to estimate the position and orientation of an exposed part, then passes a candidate pose to the robot system for clearance and path checks.
Published supply evidence
TransferTech Epic Eye Laser L V2S lists a recommended working distance of 1.2–3.0 m. Mech-Mind documents a sheet-metal picking example that separates pose recognition from robot path planning.
What to verify
Check coverage at both the top and bottom of the bin, plus occlusion, pick success and sustained cycle time with actual parts.
Review this application →
Transfer a presented part
  1. Capture the tray

    Overhead vision camera above a PCB tray

    See which pocket is occupied.

  2. Locate the board

    Illustrative camera view showing PCB position and angle

    Estimate position and angle from the board outline.

  3. Transfer the board

    Robot gripper transferring a board from a tray to a fixture

    Adjust the pickup pose and move the board.

Illustrative sequence · The vision output guides the robot action.
How it works
An overhead image shows which tray pocket is occupied and where the board sits in it. Use the visible outline or reference features to estimate position and angle before moving it.
Published supply evidence
Percipio PS800-E1 has a published 0.2–1.0 m working range and a pick-and-place application reference. The PCB image is a conceptual tray-to-fixture workflow. Board-specific tooling and ESD compatibility require qualification.
What to verify
Verify gripper contact and ESD requirements, board damage, transfer time and repeatability across revisions.
Review this application →
Correct pose at placement
  1. Capture the fixture

    Vision camera looking down at a part and locating fixture

    Locate the part relative to the fixture.

  2. Calculate the offset

    Illustrative overlay of measured and target outlines

    Compare measured and target positions.

  3. Correct the placement

    Robot arm placing a bracket at a locating fixture

    Update the robot pose and verify seating.

Illustrative sequence · The vision output guides the robot action.
How it works
Measure the part’s position and angle against fixture datums. After camera–robot calibration, use that offset to update the target pose before placing the part.
Published supply evidence
Mech-Mind documents 3D pose recognition and robot path planning using Mech-Vision and Mech-Viz. These are building blocks for the workflow. The locating fixture shown is an engineering illustration.
What to verify
Check robot and controller support, calibration drift, actual seating tolerance and cycle time. Confirm who owns integration and acceptance testing.
Review this application →

Published specifications support initial selection. Confirm application performance, lead time, local support and integration scope with the supplier or integrator after reviewing your parts and site requirements.

Compare the decision points in your cell

Start with how the part arrives and what the receiving station needs from the robot.

Pick from loose supply

Start here
Incoming parts overlap and their pose is unknown.
Screen out if
There is no usable view or accessible grasp near bin walls or at the bottom.
First acceptance check
Run full-to-nearly-empty bins. Record successful picks, retries and operator interventions.

Transfer a presented part

Start here
A tray or fixture presents a part within a known range of positions.
Screen out if
Allowed contact areas, part variation or static-control requirements rule out the proposed gripper.
First acceptance check
Confirm contact points and verify repeatable pick and transfer without handling damage.

Correct pose at placement

Start here
The receiving fixture or process needs tighter pose control than the incoming presentation gives.
Screen out if
Camera and robot cannot be calibrated to a stable datum or placement cannot be checked.
First acceptance check
Measure deviation at the fixture and confirm seating, clearance and cycle time.

How a robot cell gets from an image to a verified placement

Separate the camera view, the coordinate transform, the grasp and the handoff. A strong point cloud is useful only when these stages work together.

  1. Capture the working volume

    Place the camera so it can see relevant parts from the top layer down to the bin floor. Check what the wall and other parts occlude.

    What we need to know Bin dimensions, mounting height, working distance and field of view at both fill extremes.

  2. Locate and transform the pose

    Estimate a pick pose from the image or point cloud, then transform it into robot coordinates using hand–eye calibration. Define the response when no valid pose is available.

    What we need to know Robot/controller model, camera mounting arrangement, calibration and interface.

  3. Plan and execute the grasp

    Check gripper contact and clearance before commanding the robot. Agree when the cell can retry, rescan or ask for operator help.

    What we need to know Part weight, allowed contact surfaces, obstacle clearance and recovery behavior.

  4. Place and verify

    Align the part to the receiving fixture. Verify that it seats in the required pose and track failed placements or manual interventions over repeated cycles.

    What we need to know Fixture datum, placement tolerance, success rate and sustained cycle time.

The first question we would ask

How far is the camera from the highest part and the bin floor? The screened cameras cover different working ranges. Mech-Mind’s published sheet-metal example also separates pose estimation from robot path planning, so we would test visibility, grasp and clearance as distinct risks.

See the example review →

How we turn supplier research into a shortlist

Example: we compare two 3D cameras against the distance from the mount to the highest part and the bin floor. This screens out candidates before you spend time on a demo or quote.

0.7–0.95 m at the proposed mount

Mech-Eye PRO S-GL stays within its published 0.5–1.0 m range. TransferTech Laser L V2S starts at 1.2 m, so it falls outside its recommended range at this mount.

1.5–2.1 m at the proposed mount

TransferTech Laser L V2S stays within its published 1.2–3.0 m range. PRO S-GL is beyond its published range. The actual bin footprint still needs to fit the field of view.

Before a shortlist becomes a recommendation

Check coverage at both fill heights with representative parts. Confirm the robot interface, calibration setup, lead time and local support. Verify actual pick and placement on samples.

Distances above are illustrative mounting scenarios. Ranges are published model specifications. Results and delivery are confirmed for each application.

Share your bin, part and robot →

Example review · feeding machined parts from a deep bin

Overlapping metal parts must be taken from a bin and placed into a fixed nest. Reflective surfaces, clearance near the walls and remaining parts at the bottom are the main unknowns.

Example: A camera can see the part at the top of a bin and still miss pieces at the bottom. Check both positions before choosing the camera.

Check imaging coverage

The Mech-Eye PRO S-GL lists a 0.5–1.0 m working distance and an 800 × 450 mm field of view at 1.0 m. TransferTech’s Laser L V2S lists 1.2–3.0 m and 1,120 × 1,030 mm at 1.2 m. Start with the mounting height and the coverage needed at the highest part and the bin bottom.

What to verify Verify usable depth coverage at both fill extremes, including occluded and reflective parts.

Check grasp and path independently

Mech-Mind’s sheet-metal example separates pose recognition from robot path planning. Gripper contact and access depend on the actual part geometry.

What to verify Run a representative mix of orientations and record success, retries and wall clearance.

Check placement and exception handling

Define how the nest locates the part and how the cell confirms seating. Agree what the cell does when the next pick cannot be made.

What to verify Measure seating repeatability, operator interventions and sustained cycle time.

Suggested next step

Request a sample-based feasibility test when camera coverage, robot reach and candidate gripping surfaces have been screened.

Request a handling review →

What to ask for in a comparable quote

Supply boundary

Separate the 3D camera and software, robot and gripper, fixture, controls, guarding and line integration. Identify who owns calibration and ongoing support.

Comparable trials

Share the same parts, fill states or tray variants, required orientation, robot model and target cycle time with each bidder.

Acceptance at the cell

Agree how to count retries, empty picks, placement failures and manual intervention. Confirm delivery, commissioning and local service responsibilities.

How we can help

Choose the level of support that fits your decision.

Supplier sourcing

Screened supplier pool and industry contacts

We match your part, process and qualification needs against our screened pool of Chinese suppliers and industry contacts. We explain why candidates fit, flag gaps and help with introductions or quote requests where appropriate.

Planning estimate: 2–3 business days for an initial sourcing reply once we have enough detail to assess the request.

Technical advisory and China visits

Research, validation and on-site discussions

We help you compare technology routes, assess the Chinese supplier market and decide what to validate before committing. If an in-person visit would help, we can shortlist sites, prepare a technical agenda and coordinate supplier meetings or factory visits, subject to availability.

Planning estimate: 5–10 business days for a focused advisory report after scope and inputs are agreed. We confirm the delivery date before you commission work.

Share your project and we’ll confirm the scope, timing and any fees before work begins. Equipment, samples, travel and third-party testing are quoted separately where applicable.

Discuss your project →

Tell us what the robot needs to move

Share your application and work email. Our team will review the details and follow up with the next questions or evaluation steps.

See a sample review →

What you can send us

Start with the part, where it comes from and where it must go. We can identify the first vision or handling test and the camera, robot and integration questions that remain.

Add dimensions or timing (optional)
Contact details
Add your name or company (optional)

We use your details to respond. We’ll ask before sharing your project with prospective suppliers.

Questions before you start

What information is useful for the first review?

Start with the part, how it arrives and where it must go. Add the robot model, preferred gripper or target cycle time if known.

Can you assess a camera separately from a complete cell?

Yes. Camera screening covers usable depth and field of view. A cell review also considers the gripper, robot path, fixture and exception handling.

Who handles sample tests and integration?

The first review can identify the work to evaluate. Test access, robot compatibility and integration responsibility are confirmed with the relevant company for each application.

Leave a message

We use your details to respond. We’ll ask before sharing your project with prospective suppliers.