Artificial Intelligence means many different things
To us at SICK it means something very special. We see it as the key to enter a new era of Sensor Intelligence. It means the possibility to solve more demanding tasks. To quickly adapt to changing conditions. To easily recognize patterns more quickly and reliably than ever before. Valuable data is collected by our sensors, interpreted by our algorithms. So you can focus on the big picture, optimize your workflows, and make efficient use of your resources.
Unlocking the full potential of Intelligent Sensors with deep learning solutions
Deep learning is subset of artificial intelligence that enables computers to mimic human decision-making and problem-solving. Our deep learning solutions help you to be more precise, more flexible and ultimately more successful. We want to give you the key to solve your technical challenges.
Get inspired by our application examples
Automated program change at the bottle cleaning machine
Bottle-washer machines in the beverage industry need to know if they clean new or returned glass. Selecting the appropriate program has previously been done manually to prevent potential downtime. Now, by using the InspectorP62x 2D vision camera combined with the Intelligent Inspection SensorApp, bottles are easily classified into new or used glass in the infeed zone of the cleaning machine.
High yield and automated wood
Naturally and uniquely grown objects are difficult to classify with traditional rule-based machine vision. Deep learning technology is well-suited for classifying the annual ring structure in the wood industry. With the InspectorP62x 2D vision camera combined with the Intelligent Inspection SensorApp, orientating of wood boards allows an automated process with high yield.
Quality control in window blinds production
Window blinds are filled with foam to reduce noise caused by things like rain. The expansion of the foam during the manufacturing process can be incomplete, which makes a visual inspection of the cut surfaces necessary. With dStudio and some representative teaching images, it is easy to prepare a neural network that will enable a SICK sensor for that task. The software learns in a human-like way how to autonomously distinguish between a homogeneous foam filling and an incomplete filling at the cut surface.
Quality control of shiny objects in product assembly
Highly reflective material is a true inspection challenge for traditional rule-based machine vision. Deep Learning is suitable for these kind of applications where the objects are shiny with some natural variances. Parts to be mounted can be missing or misplaced, such as screw threads when mounting a product. Classification of mounted parts with a shiny surface can be done by using an InspectorP62x camera for image acquisition together with the Intelligence Inspection SensorApp.
Accurate orientation of fish in production
Objects with slight differences within one class are difficult to differentiate with traditional rule-based machine vision. Deep Learning is well suited to classify the orientation of fish. With the InspectorP62x 2D vision camera combined with the Intelligent Inspection SensorApp, knowing the orientation of fish on the production line allows for accurate and automated processing.
Create your own SensorApp with SICK AppSpace
The SICK AppSpace eco-system gives you the possibility to create your own SensorApp that exactly meets your requirements. You can, for example, create a SensorApp that will look for foreign objects in a homogeneous bulk material. Deep Learning will not only allow you to solve such a specific use case, it also enables your end users to simply readapt your solution to any similar application.
Solder joint inspection
The inspection of solder joints can be challenging due to surface reflexes combined with a large variance in their visual appearance. If you can explain what to look for on sample images, Deep Learning will not only deliver high detection accuracies, but also significantly ease and speed up your solution development. With dStudio and representative teaching images, it is easy to prepare a neural network that will enable a SICK sensor for that task.