Research On Defect Location And 3D Segmentation Technology Based On Industrial CT Image

Industrial CT is an effective detection technology for workpiece inspection and quality evaluation, which is able to get internal structure and defect information of product nondestructively. Now, with the development of CT reconstruction and visualization technology, the workpiece could be viewed not only in 2D space, but also in 3D space. Thus,defect location and 3D image segmentation technology should be study and implemented more efficiently and accurately to meet the higher requirements. So that technicians can analysis and measure defects more directly by 3D view. This paper mainly researched the automatic defects location and 3D image segmentation technology of industrial CT image,and developed the corresponding analysis software. The research results has considerable practical value in quality inspection and maintenance for industrial products.The main work of paper can be summarized as follows:1. For 2D CT slice images, a defect automatic location algorithm based on fractal dimension is proposed combining fractal theory. Firstly, fractal dimension is calculated for Sub-Block images. Then, the defect area can be located by setting a fractal dimension threshold. Experiments show that the algorithm is effectively and accurately on defects location.2. For 3D image slice sequences, this paper researched a 3D defects segmentation algorithm which is based on morphological and Otsu, to segment defects of slices layer-by-layer and extract the 3D defect data accurately. The effectiveness of the proposed method is verified by segmentation experiment of the simulation solid-propellant rocket engine slice sequences.3. For 3D volume data, an improved method for 3D image segmentation is presented based on the modified k-means, and which can generate initial clustering centersautomatically and avoided artificial selection in traditional k-means clustering algorithm. As the algorithm takes full advantage of the whole information in 3D, the volume data of defects can be segmented and obtained more accurately.