Machine vision · Guidance
When the part does not always arrive in the same position, the robot needs to see it. The cell combines camera, calibration, lighting, algorithm and gripper. Feasibility depends on part variation, cycle, placement accuracy and station safety.
Part location and orientation so the robot can pick, place or assemble without rigid presentation.
What it solves
Comparison
The decision depends on whether the part is planar, stacked, moving or requires full orientation.
| Criterion | 2D | 3D | 3D + bin picking | Conveyor tracking |
|---|---|---|---|---|
| Best for | Flat parts and known position | Height, volume and orientation | Stacked or random parts | Moving part |
| Data required | Contrast, color or shape | Point cloud or profile | Cloud, separation and feasible grasp | Encoder, trigger and speed |
| Limit | Limited height and occlusion | Calibration and reflective surfaces | Cycle and collision risk | Latency and synchronization |
In detail
Camera and robot must share the same coordinate system. Calibration is validated across the whole work volume, not just the center.
Vision finds the part; the gripper decides whether it can be picked. Shape, weight, surface and clearance define the end effector and grasp strategy.
The real cycle includes seeing, thinking, moving, picking, verifying and placing. Occlusion and station safety may require changing the camera or layout.
Vision integrates with robot, PLC, conveyor and safety system. Retry and reject logic is defined before commissioning.
Industries
Related guides
Frequently asked questions
When height, orientation or overlap matter. If the part is flat and always in the same plane, 2D may be enough.
No. The part must be separable, graspable and removable without collision. A trial with the real part confirms feasibility.
With calibration points and validation across the work volume. Calibration is verified with the real gripper and part, and documented so it can be repeated.
No. Vision solves uncertainty; it does not fix feeding that cannot separate or present the part. A better-designed tray is often cheaper than a camera.
The cell can reject it, retry or request intervention. The logic is defined with the customer and tested with borderline cases.
Yes, with an encoder and synchronization logic. Latency and conveyor speed define the time available to process the image.
Part, feeding, weight, tolerance, cycle, robot and gripper. With that, camera, optics and algorithm are defined and tested with real parts.
Vision integrates with the existing robot and PLC when interfaces allow it. Compatibility is validated during assessment.
Next step · Gist Point
Send photos or a video and we will define whether a vision trial, an assessment or a complete cell makes sense.