Image analysis of complex microstructures by texture analysis and correlation with properties by neural networks

By characterising the microstructure, quantitative image analysis allows to draw conclusions on the mechanical properties of materials. On fine micros tructures with low contrast, e. g. of hardened steels, texture analysis has to be applied for quantification. Feeding texture parameters according to Haralick into a trained neural network, a correlation between the microstru cture and the hardness of the steels C45 and 100Cr6 can be achieved.