Computer vision-based technology has enormous potential for quality control, such as for verifying that cameras coming off the line have all components and assembled correctly. The TARGET build state is often kept in CAD data, which is why processes are needed that detect the ACTUAL build state and register it to the CAD data.
We use augmented reality processes that can automatically detect differences between TARGET and ACTUAL in real time. The processes can be easily adapted to a broad range of different product configurations, they require no image-based training since our processes rely on the design data and they can even be set up during the production planning process. This allows the implementation of versatile inspection processes that can be easily adapted to any number of product variants. Specimens can be recorded with either 2D camera arrays or 3D inspection camera systems.
Augmented reality-based quality control can also be used for mobile inspection systems, wherein the inspection engineer uses the camera on a tablet to photograph the specimen and register it to the CAD model in real time. This allows the inspection engineer to identify and document variances between the CAD model and the real model.
Our augmented reality processes use model-based 3D object tracking without the need for markers. Our tracking solution is robust, accurate and easy to integrate. It is also very effective in inconsistent lighting conditions. A monitor displays the CAD model superimposed on the camera image and differences are highlighted.
In our sample scenario, a sheet metal test part is recorded by the cameras. When the test object is placed on the table, the part is detected and its orientation recorded. The augmented reality visualization identifies and documents variances between the part and the CAD model. 3D cameras are also used to detect deformations in the parts.
The demonstrator has tremendous potential that opens augmented reality to industrial applications. In the context of industry 4.0 in particular, this technology will be used to continuously compare 2D simulations to real machine configurations.