Today's 3D cameras are not able to see transparent parts, which leads to challenges in relation to automation of several applications, for example:
- Full automation of several types of logistics and warehousing operations is not achievable. This limits the market as potential buyers of automation solutions require solution providers to handle all parts in the warehouse, both non-transparent and transparent parts.
- Within manufacturing robot picking is limited regarding what problems that can be solved and thus limiting the market opportunities.
By combining the latest research in 3D machine vision and deep learning, in an innovative combination with new hardware design, Zivid will develop solutions that enable 3D imaging of highly challenging transparent objects. The most important R&D challenges we expect to meet are:
- the development of new imaging methods that handle the physics when light interacts with transparent parts
- build enough representative training data for deep learning
- develop new 3D hardware that supports and enables the imaging of transparent parts.
The starting point for the project is the next generation Zivid camera, a 3D color camera that is particularly suitable for robotics and which has a number of new unique features that we believe can be utilized in an innovative way for imaging difficult transparent parts.
Through DeepStruct, Zivid will work together with Norway's leading R&D environment within 3D image processing and machine learning at SINTEF. This will provide both world-leading research; and form the basis for a continued success story for 3D camera technology developed in Norway.