DETECTION OF DEFECTS IN GROOVE TEXTURES OF HONED SURFACES

The automatic assessment of the texture of honed surfaces of cylinder bores based on microscopic grey level images is a demanding task. Alth ough for some of the problems arising in this context, solutions are a lready given in literature [6, 7, 9], there remains a lot of work to b e done due to the complexity of honing textures and the high quality d emands made by combustion engine manufacturers. This paper deals with the special task of automatically detecting undesired defects like: fo lded metal, groove interrupts, smudgy groove edges etc. in honing text ures. Therefore, two different image processing algorithms are present ed. The first one searches for defects locally, whereas the second alg orithm aims at detecting defective grooves in their entirety by exploi ting the fact that, if a groove is affected, then usually several defe cts occur within the very same groove. The signal theory necessary to understand the algorithms is shortly presented. It is demonstrated tha t both approaches deliver reliable detection results for real honing t extures. (C) 1997 Published by Elsevier Science Ltd.