Image processing and classification of metal fracture surface

In this paper, a method of the image processing of the metal fracture for eliminating the noise, detecting the edges of the image and classification of the typical morphology of the metal fracture surface by obtaining the images’ character based on the mathematical morphology will be described. The methods of the normalized pattern spectrum have unchangeable characteristic for rotating, zooming or moving the images. Then the images can be classified by using artificial neural network. In this paper, 60 images of corrosion fatigue fracture, stream design fracture and intergranular cracking fracture are used as a study samples and 48 images are used as a test samples. The data of the experiment explain the feasibility of the method and analyze the cause of the error.