I can see what I'm doing - easy for humans, a challenge for robots. Image processing is playing a key role in anchoring robotics applications in more and more areas of automation. Vision sensors such as the Baumer VeriSens XF900 and XC900 models can provide support here. They combine a complete image processing system in one housing and can be easily parameterized. This makes image processing much easier for users.

What are the benefits of image processing for robots?
Robots orient themselves to taught-in, fixed waypoints and follow them in sequence. Image processing extends this functionality enormously. A simple vision sensor already enables the robot to reliably identify objects. At the same time or alternatively, quality control can also be implemented, which the robot supports by moving the sensor to defined positions. By using the data in the robot program, objects can then also be sorted automatically.

The "supreme discipline", however, is controlling the robot itself, in which image processing determines the position of objects, transfers it to the robot and thus enables it to grasp freely beyond previously fixed waypoints. Objects can be positioned in any orientation on a surface. Position, rotation and optionally other data such as their location are determined and transferred by the image processing system.

Universal Robots uses so-called caps for certified accessories, similar to the "app concept" familiar from smartphones: software plug-ins that make attachments such as a VeriSens vision sensor usable in the Universal Robots programming environment PolyScope.

The image processing task itself is parameterized via the VeriSens Application Suite - independently of the cobot in the most suitable environment. The functions of the VeriSens URCap are generic and thus address all conceivable applications, including the accompanying or stationary arrangement of the vision sensor.

In addition to the installation routine shown in the URCap, only two additional "nodes" (commands) are required to program the Universal Robot, with which image processing is incorporated into the robot programming. For object identification or quality control, for example, a single node is sufficient to trigger an image processing job on the VeriSens in the robot program and make the results available as a variable in the program sequence for decision-making. This means that the cobot can already sort objects.

The second node is added for image-based gripping, which supplements the fixed waypoints with dynamic, image-based ones. An optional wizard specially created for this purpose helps with the simple implementation of pick-and-place applications.

The highlight with automatic calibration
Robots and vision sensors work in their own coordinate systems, which only becomes functionally relevant when object positions are transferred from the VeriSens to the cobot. The coordinates of the vision sensor must be converted to correspond to the robot coordinate system. Until now, the necessary coordinate transformation was determined as a "hand-eye calibration" by manually positioning the cobot several times with a probe tip on a special calibration target. A procedure specified by the manufacturer is run through in many individual steps. This is laborious and also inaccurate and error-prone due to the manual intervention in terms of human control of the probe tip.