An Intelligent System for Analyzing Welding Defects using Image Retrieval Techniques

The development of new approaches in image processing and retrieval provides several opportunities in supporting in different
domains. The group of welding engineers frequently needs to conduct visual inspections to assess the quality of welding products.
It is investigated, if this process can be supported by different kinds of software. Techniques from a generic CBIR system have
been successfully used to cluster welding photographs according to the severeness of visual faults. Similarity algorithms were
used to automatically spot faults, such as cracks and gas pores.