Camera-guided robotics and pick-and-place

A robot can move very precisely to a known position. It gets harder when the part does not lie in exactly the same place every time.

This is where the vision system provides orientation: it detects the part, determines position and rotation, and makes that information available to the robot controller.

Robots can then pick up, align or place workpieces even when their position varies.

Discuss your application

Typical applications

pick-and-place
picking parts off conveyor belts
removing workpieces from trays
feeding components in the correct orientation
inserting parts into fixtures
loading and unloading machines
determining assembly positions
correcting gripper position
sorting parts
determining position ahead of machining processes

2D or 3D?

Not every robot application needs 3D.

2D vision: Particularly suitable when the parts lie on a defined plane and mainly X position, Y position and rotation have to be determined.
3D vision: Becomes interesting when height information, differing Z positions, tilted objects or randomly oriented parts in bins also have to be taken into account.

For classic pick-and-place of flat-lying parts, 2D is often the simpler and more economical solution.

Stationary camera or camera on the robot?

Both concepts are possible.

Stationary camera: The camera monitors a fixed area and transmits the position of the workpiece.
Eye-in-hand: The camera sits on the robot and moves with the arm.

Which setup makes more sense depends on field of view, accuracy, range of movement and process.

Calibration is decisive

For a position detected in the image to become a robot movement, camera and robot coordinates have to be calibrated to each other.

In robot vision projects we therefore look beyond camera and lens at:

working area
required accuracy
robot
interface
camera position
gripper
cycle time
lighting
calibration

This information helps us specify the system

Example values - the robot make and the required gripping accuracy matter most.

Working area: 400 × 300 mm
Gripping accuracy: ± 0.3 mm
Part position: flat, any rotation
Robot: 6-axis, PROFINET
Cycle time: 2 s per part

Suitable product groups

Locating parts by camera or building a pick-and-place cell?

Tell us the working area, the accuracy you need and the robot - that is enough to narrow down a suitable setup.

Discuss your robot vision application

Frequently asked questions

Do I always need 3D for pick-and-place?

No. If the parts lie on a defined plane and only X, Y and rotation are needed, 2D is usually simpler and more economical. 3D becomes relevant with height differences or randomly oriented parts in a bin.

Should the camera be stationary or mounted on the robot?

Both work. A stationary camera monitors a fixed area; an eye-in-hand camera moves with the robot. The choice depends on field of view, accuracy and range of movement.

What is hand-eye calibration?

The mapping between camera and robot coordinates. Without it, a position detected in the image cannot be translated into a correct robot movement.