抄録
The software which enables us to classify six typically different fracture surface morphologies, such as striations, equiaxed dimples, elongated dimples, transgranular failures, intergranular failures, and cleavage failures, has been developed based upon computer image processing and pattern recognition. The texture analyses used here are the Fourier power spectrum, generalized co-occurrence matrices, and three-dimensional fracture surface topographies. Fourier power spectrum analyses are very powerful for recognizing striations and intergranular failures, whereas a generalized co-occurrence matrix and fracture surface area ratio are suitable for cleavage failures. We also point out that the membership values based upon a soft clustering method, a fuzzy C-means algorithm, allow the system an uncertain fracture surface. This increases the precision of recognition.
