Here are answers to some of the most common questions we receive.
With the standard settings, the AI model can detect when the parts have slight deviations in dimensions or missing features. If requested we can train the AI models in a custom way so that they can detect dimensional deviations few mm tolerances and if there are missing features on the part, the missing features need to be specified in advance.
We usually use 6, 8 or 12mm lenses depending on the application. These produce approximately 20, 30 and 40 degrees vertical FoV respectively.
There's really no limitation in terms of range, as long as we increase the stereo separation of the cameras to accommodate larger distances. Our standard rig works well from 30cm to 1 meter. The important thing to make sure is that the part that is being located is not too small in the images. So locating a 1 meter wide object at 3 meters distance is fine, but a 5cm object at 3 meters will force us to zoom in on the image and sacrifice the FoV.
The standard cameras for the Cambrian system are manufactured IDS Imaging Development Systems GmbH and the images can be accessed using IDS software or API provided by them. The Cambrian system also maintains a log of all the raw images than are accessible on the Cambrian PC.
This is up to you. We suggest that the box should be as shallow as possible. It should also be reachable by the Robot Arm in every location.
The Angles at which the parts are detected can be set up before training. Normally each Pick point has 45° of x and y, and 90° or z rotation. This can be changed.
This depends on your object size and the accuracy you want to achieve. Generally you can say that the closer you are to the part, the more accurate you are. As a general feeling I would say that for 1mm sized parts you want to be around 300mm from the part. For parts 200mm sized you can be around 600mm away. The height for which you can detect a part is set in training and can be seen on the Right side of the "AI '' Tab in the Cambrian Software, when an AI Model is selected.
Online support is available during business hours (German Business hours).
Most other vision systems take a few seconds to make a scan, process the data and subsequently deliver the result. Because they're so slow (and sometimes bulky too), they are usually mounted in a stationary position above the bin so the system can scan while the robot is moving and essentially "hide" the system's cycle time. On the other hand, Cambrian Vision is pretty quick. In a well lit room, the whole capture+data processing+reply takes ~200ms. Mounting the cameras on the robot, Cambrian enables a lot of functionality and flexibility that wouldn't be possible otherwise. Having said that, we might enable this as a feature at some point because it is technically possible to do that. Please send us a request if your application requires this feature.
Cambrian PickEditor can load at the moment .obj and .fbx file formats. More common file formats such as: STEP(.stp, .step) IGES (.igs, .iges) Solidworks (.sdprt) Catia (.catpart) PTC Creo (.prt) STL (.stl) can be received and work well, but these need to be converted to .obj or .fbx format. We recommend using https://www.opencascade.com/products/cad-assistant/ to convert these files. When sending files to us, please send in native format that you have - so we can check that the conversion worked properly.
Please get in touch using the Contact us form
Please get in touch using the Contact us form
Please use the form below to contact us.
Let’s discuss how we can help you
Send parts to us and we’ll
send back a video
Contact us for all product
enquiries
Contact us to schedule an
online product
demonstration