Image processing is still considered by many users to be a field that requires a great deal of specialist knowledge. Especially when using collaborative robots (cobots), it is important that the robot can 'perceive' its environment in order to enable safe and efficient interaction with humans. In addition, image processing is indispensable when robots have to recognize objects that they are to manipulate - in the sense of moving, assembling, sorting, picking or packing.

However, image processing systems do not necessarily have to be highly complex and require a great deal of expertise. The basic structure of image processing is always similar and consists of three steps: Capture image, evaluate and output results. The concept of smart vision sensors is based on implementing the three steps in a compact device. One example of this is the 'O2D500' from ifm.

Robot sensor calibration with a click

With the calibration sheet, the vision sensor can be calibrated to the robot's coordinate system with just a few clicks.

© ifm

When using a vision sensor with a robot, one of the most important tasks is to calibrate the image captured by the sensor and the robot's coordinate system to each other. To make this task as simple as possible, the vision sensor from ifm uses the 'Vision Assistant'. When developing this software, the focus was on ensuring that it can also be operated by users who are not proven specialists in image processing. The tool therefore contains various image processing algorithms that can be used without the need for programming. This also applies to the robot sensor calibration, which is based on a so-called marker calibration. It works largely automatically with just a few clicks. To do this, the user can print out a calibration sheet from the software and then place it in the robot's working area and in the vision sensor's field of view. Once the focus and exposure have been set - this is also done automatically with just a few clicks in the software - the robot's tool center point must be placed on the four markings one after the other. The user enters the coordinates in the corresponding fields in a table in the 'Vision Assistant'. In the next step, the data is transferred by clicking on the Teach button. Finally, up to 16 images are taken from the calibration sheet, each in slightly different positions, in order to improve the calibration. The software displays a quality indicator: Calibration is complete when this is at least 85 %. If required, a Z offset can be specified. To check the calibration once again, the user can measure the length of a line, which is also printed on the sheet, in the software. If this matches the actual length, the vision sensor is ready for use.

Combination with other functions

The author: Dr. Jörg Lantzsch is the owner of the Dr. Lantzsch agency in Wiesbaden.

© ifm

A helpful feature of this type of robot-sensor calibration is that the sensor adapts to the robot's coordinate system. This eliminates the need for coordinate transformation in the robot program, which is always a potential source of error. To further simplify the application, there are example programs for various robot manufacturers that users can work with directly. The syntax can also be created for other robot types with just a few clicks. The communication modules from the Vision Assistant to the robot are also preconfigured, with communication between the robot and vision sensor taking place via TCP.

Other functions such as contour and object analysis are already programmed in the Vision Assistant and can be combined with the robot functionality.