Abstract
Porosity structure known as void often appear in casting objects and molded object. They often cause structural weakening of the object and it is important to analyze their shapes and distribution in order to improve manufacturing process. This paper presents an efficient method for extracting void structure of mechanical objects from CT images. Our method specifically enables to extract open voids distributed on the surface of the object based on mathematical morphology. In order to handle large CT images, we introduce out-of-core method for mathematical morphology on GPGPU with low memory consumption. In addition, we also developed several tools for analyzing void distribution for non-destructive evaluation.
