Research Of Some Key Algorithms In Metallographic Image Analysis System

Metallographic image analysis is one of the most important means to test Metallographic sample, key algorithms in metallographic image analysis are image segmentation. This method includes objects detection and segmentation in conglutination objects.Combing metallographic image analysis with computer assistant measure , we have developed the Metallographic image analysis system successfully. This system can make many functions such as recording images and analyzing those images quantitatively:Record the Metallographic images and all kinds of parameters of the images.Pre-processing for the recording Metallographic images, pick-up of objects and segmentation in conglutination objects.Measure, compute the basic characteristic parameters of the image of processed.,Compared with the existing segmentation methods in metallographic analysis, this system is robust, adaptive, and unsupervised.In this paper, we just make researches on some key techniques in Metallographic image analysis. According to the characteristic of metallographic images and doing many experiments carefully, we present and realize the following algorithms or methods.1) According to the geometrical diversity of the metallographic objects, this thesis presents adaptive test and methods to discuss metallographic images and improves the anti-watershed algorithm to segment the conglutination objects.2) According to the geometrical diversity of the metallographic objects, applies a polygonal approximation to every object where several particles may touch each other.3) Improves the methods that using classifies concave points on object boundaries to segment the different classes based on its angle and lengths of 2-vertex lines.