An Image Processing Approach for Improving the Recognition of Cluster-like Spheroidized Carbides

This paper was mainly applied to image identification of metallographic structure of carbon steel. Though metallographic image identification is now needed by industry, it is rarely discussed in literature due to its industrial characteristics, let alone the theory of identifying complex structures. The identification of metallographic structure of common carbon steel is mostly carried out manually, which is mainly plagued by empiricism and subjective identification. This paper intended to calculate the percentage of spheroidized carbide in metallography. However, the distribution of carbides is affected by the insufficient heating process. For example, low heating temperature or short holding time will result in carbide connection, which leads to the reduction of the accuracy rate in calculating the spheroidization rate of carbide. However, the algorithm proposed in this paper mainly strengthens the accuracy rate of carbide cutting, and the connected carbide is morphologically cut to improve the identification accuracy rate. For carbide cutting, it is carried out in two stages. First, all disconnected components are cut by using the connected components, and then morphological erosion and expansion calculus are carried out for all carbides to cut connected carbides.