Automatic design of classiffcation systems for visual quality control of metallic surface

Daniel Garten, Katharina Anding, Galina Polte, Konstantin Trambitckii

Abstract

In this paper we propose a new method for the automatic design of classification systems for the visual inspection of metallic surfaces. Feature selection and selection of a suitable classifier as well as its optimization are done fully automatically on the base of a set of representative images for all quality respective defective classes to distinguish. Texture and colour features, which are most suitable for discrimination between different quality classes, are also explained. Furthermore, we present a flexible software architecture based on Windows Workflow Foundation and Windows Communication Foundation for the separation of image acquisition and classification. With this scheme a process architecture with many thin clients for image acquisition and one high-capacity evaluation system can be built up.

Download

IMEKO-TC10-2016-005.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings