Oberflächenprüfung mit neuronalen Netzen

Surface Inspection has a broad variety of potential applications in computer vision. Nevertheless there are only few inspection systems that have become commercial products. One reason is the very rigid structure of the algorithms that are mostly designed for one special inspection task. Neural networks provide a mechanism to get the image processing more flexible and thus the developing time for a inspection system shorter. An introduction to an image processing and pattern classification concept is given and the possibilities and boundaries of the concept are discussed.