The Method of Edge Detection Based on Interval Type-2 FCM

The traditional clustering method, in some industrial fields, could not produce an ideal effect on the edge feature extraction, such as the crystal edge extraction in the process of dealing with the growth of single-crystal silicon. Based on this point, this thesis puts forward The Method of Edge Detection based on Interval Type-2 FCM. This method extends the traditional FCM(Fuzzy C-Means) from Type -1 fuzzy set to Type -2 fuzzy set and adopts Gaussian kernel to replace the original Euclidean distance in order to make up for the disadvantages of the traditional clustering algorithm FCM and reduce the possibility that the influence of similar features is magnified. Use the improved method on IRIS test set and Lena test pattern respectively to make a test and the result of the test shows that improved Interval Type-2 achieves obvious improvement on classification and the effect of image segmentation. A good result of segmentation could be gotten when making use of improved method of edge detection based on Interval Type-2 FCM to make feature extraction of single-crystal silicon in the phase of seeding and isometric.