Inspecting surfaces with machine learning
In order to reuse components, such as those of an engine, their condition must be analyzed in a complex manner. Only a precise inspection of the surface and condition provides information on whether the component can be used again. In times of scarce raw materials, the efficient performance of such analyses is becoming increasingly important.
State of the Art
A common approach is to compare individual images or images from line scan cameras with a known, intact image of the object to be analyzed, often a 3D (CAD) model. This procedure places high demands on calibration and requires the position of the component in space to be known. If the object deviates minimally from the model, for example because it is a different version or a component with signs of use, there is no meaningful result.
Technology
Researchers at the Institute for Industrial Information Technology (IIIT) at KIT have developed a method that analyzes objects based on video data and machine learning. Based on previous inspections, a suitable trajectory for the moving camera is automatically determined so that the relevant object areas are clearly visible. The industrial 3D camera moves around the illuminated object and captures it as a video stream. For larger objects, the use of a drone is conceivable. A 2D representation of the surface is created from the video data. A trained neural network knows different structural shapes and states, so it can analyze large variances of models and states of a component in real time. To speed things up, the system operates at two resolutions: In the low resolution, the relevant surface areas are detected. These are then inspected for defects using high-resolution images.
Advantages
The automatic, intelligent optical visual inspection enables components with a large variance of model expression and states to be analyzed efficiently. The use of video data ensures efficiency and reduced susceptibility to defects, as the relevant surfaces are viewed together rather than as individual images. The method therefore works independently of the position of the objects relative to the camera.
Options for companies
KIT is looking for partners to bring this new type of measurement technology into application, e.g., companies that analyze components for the purpose of reuse or system integrators that establish and support the necessary technologies.
Your contact person for this offer
Innovation Manager Mobility and Information Karlsruhe Institute of Technology (KIT)
Innovation and Relations Management (IRM) Phone: +49 721 608-28460
Email: birgit.schulze@kit.edu