24 Jul 2024


Vision sensors identify objects without experts

Sensopart says that its AI-based vision sensors simplify image processing and can be set up in just a few steps

The German sensor-maker SensoPart has developed an image-processing system which, it says, can detect and differentiate items reliably without users needing to be machine vision experts. The Visor Object AI system implements an AI algorithm in the vision sensor, thus avoiding the need for network or cloud connections.

The system learns distinguishing characteristics of an object using as few as five images of the object. It can differentiate between variants even for flexible, shape-changing objects such as spiral springs or plastic bags. It is not affected by fluctuations between batches, contamination, reflections or varying 3D orientations. After being taught using a few mouse clicks, the system can recognise objects and assign them to different classes.

For presence checks, components can be rated simply as “good” or “bad”, or they can be divided into 200 classes. Once a classification has been taught, it works reliably without the user needing to worry about the detection rules and parameters required for traditional rules-based image processing, which relies on techniques such as pattern comparison, or contour or contrast recognition.

Compared to traditional detectors, Sensopart says that its AI vision sensors need much less set-up effort and offer increased stability. Users can save time because they do not need to create logical links between several detectors.

SensoPart:  LinkedIn