The people in white coats have come up with yet another groundbreaking development in Scorpion Vision Software. And with a great name too. 3D-MaMa is described as a 3D Pose Estimation for Random Bin Picking by Pairwise Manifold Matching.
In practical terms this enhances the existing 3D Toolset in Scorpion making it faster and easier to use. A key benefit is the recognition of an object even with a complex background. The algorithm is less sensitive to the ‘noise’ around a recognised object and this has a direct benefit for robot pick and place. Traditionally, object selection is based on segmentation of a range of images. The new algorithm searches through pairs of 2D manifolds to localise candidates for an object match.
The impressive capability of this new Scorpion Vision tool can be demonstrated simply in these images.
Scorpion has identified the parts which have been placed haphazardly (as in the case of a parts bin). The parts identified ready to pick are visualised in various colours. This clearly demonstrates the ability of the system to locate and ‘extract’ all the complexities of an object from the background and from it’s neighbours.
Time taken to process one image? 380 milliseconds, well within the cycle time of a large pick and place robot.
If you’d like to find out more about this there is an opportunity to attend a free seminar in September 2009, click here for further details.
There’s also the opportunity to attend a 2 day training course at the same time, click here for more information.
PW
Filed under: 3D Machine Vision Tagged: | 3d-MaMa, abb robot, monopose 3d, random bin picking, robot pick and place, robot vision
