On-line surface inspection with fuzzy-ASIC modules

The paper presents a system for on-line failure detection on cold-rolled steel. A special hardware system based on FPGA modules was developed to get the necessary performance. Fuzzy Pattern Classification is used as classification method. The input vector for the classification consists of several statistical features taken from the co-occurrence matrix and the difference histogram. The described system is able to detect surface failures greater than 1 mm in real-time at a strip speed up to 6 m/s. Failures are allocated to about 10 failure classes.