This work introduces a concept for a challenging classification problem of highly dynamic laser welding image sequences. Under the conditions of the industrial production a decision is required within seconds with a high selectivity. The introduced concept consists of classic elements of image processing, classification, and machine learning. The components are not optimized but as a whole, yielding a high-performance system. It also contains individually a statistical preprocessing for change detection in image sequences. The event hypotheses generated by the algorithm are classified frame-wise by an object classifier. The optimization of recognition performance is based on a feature selection by a modified sequential forward selection. We use the properties of the polynomial classifier for the efficient computation of cross validations. © Springer-Verlag Berlin Heidelberg 2003.
CITATION STYLE
Hader, S. (2003). System concept for image sequence classification in laser welding. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2781, 212–219. https://doi.org/10.1007/978-3-540-45243-0_28
Mendeley helps you to discover research relevant for your work.