Independent performance assessment of automatic surface inspection systems through integration of computer-aided external supervision procedures (CheckSIS) : final report

Aim of this project was to develop methods enabling the automatic verification of results from automatic surface inspection systems (ASIS). Therefore, new methods for computer-aided ASIS supervision were developed, employing artificial defects (AD) and synchronization of two ASIS installed at the same line. For the use of ADs finally two industrial prototypes could be implemented at the plant: Projection prototype for areascan systems at Tata Steel ; Burn-in approach for linescan systems using reference coils at Rasselstein The installed systems are able to detect decreasing system performance and thus ensure long-term ASIS stability and documented reliability. The reference coil approach is just applicable at electrolytic lines, were it is possible to switch of the coating process. To have a solution also for other plants where the use of ADs is not feasible another solution was developed to determine at least the local sharpness of the ASIS camera based on raw signal analysis. Furthermore, an online synchronization of results coming from two ASIS installed at the entry and exit section of the same line was implemented. It could be shown that ASIS classification performance can be significantly improved by combination of ASIS entry and exit results. The project results were summarized in a first official standard for ASIS calibration and supervision procedures titled: VDI/VDE/VDMA 2632 Part 4.1: Machine vision: surface inspection systems in flat steel production: stability testing. This new guideline should lead to more reliable, comparable and traceable ASIS results and finally significantly improve quality decisions based on surface defect appearance.