The CV4CE (Computer Vision for Efficient Circular Economy) project aims to develop an innovative software approach that contributes to automation and increased efficiency in the circular economy, thereby conserving resources and avoiding waste in production. At the same time, the process can contribute significantly to flexibility and cost savings in a variety of applications in the context of Industry 4.0.

In industrial circular economy and production, robotic arms are increasingly being used to grasp objects in order to analyze, sort, or feed them to the next process step. Traditionally, the robotic arm performs fixed, pre-programmed movements for this purpose. Objects that the robotic arm is to grasp must be arranged in a predefined pose (position and orientation in space), which severely limits the flexibility of the conventional approach. By using camera-guided robotics, this step can be automated and made flexible by determining the current object pose from a camera image. Current machine vision methods typically require precise knowledge of the size and shape of the objects, which is provided in the form of a 3D object model. With conventional methods, this 3D object model must be created in advance in a complex procedure. For each new object, a new 3D model must therefore be created and provided, which drives up the cost of the overall process and can make projects in production and the circular economy uneconomical.

The software approach to be developed should not require a pre-created 3D model, but instead create a 3D model from the camera image itself and simultaneously determine the poses of the object instances depicted.