Metal Fracture Sem Image Feature Extraction And Analysis

The study of metal fracture image is currently concerning the classification of images. Therefore, little research was done toward the inner characteristic of the similar image, whose characteristic has a close relationship with its life-span. Hence, to study these images with the following ways has a significant meaning:Power spectrum from 2D Fourier transformation: slide-crop a big image into small ones, obtain its Ring and Sickle characteristics after Fourier transform. Then, extract features of these small images, an ambiguous image will be distinctly described in a much clear way. And a direction field graph of the original image was also drawn.Wavelet decomposition: Db4 wavelet decomposition and energy norm L1 was employed to extract feature of Zr-4 alloy images. With the selected wavelet Db4, features at different frequency channels of life span were gotten. Then an energy change at the direction of fracture extension was studied to find the relationship between the feature and life-span.Fractal algorithm: Feature of image at different channel was extract after wavelet transform. With the data from fractal algorithm, the development of fracture expand was conducted. Fractal coefficient of decomposition images was sought by fractal algorithm, and the development of the metal fracture can be analyzed by the changes of the coefficient.Gabor transformation: In frequency field, Gabor transform has sound direction selectivity. Selecting a certain direction and picking-up corresponding feature value by this means was another choice for image research.To research the texture with these ways are a relative new subject. And it is novel to plot the fracture development with direction field. To express the texture with a clear way for material research has a significant meaning.