High precision industrial CT is an important technique for detecting of defective image. This paper is mainly involved in research of such techniques as the auto-localization, segmentation and feature extraction of defects for the solid-propellant rocket engine with data obtained from the industrial Computer Tomography. The research results of such techniques are of great significance for application, installation, maintenance, life extension and cost reduction of the solid rocket engine.First,locating the defect would correspondingly reduce the time spending. In some cases, it even only requires to locate the defect. A fast defect localization algorithm is studied in this paper, which, in combination with the fractal algorithm of morphology, conducts the defect analysis in accordance the difference of fractal dimension between defect and background and locates three-dimension engine data.The segmentation algorithm is the foundation of auto defect recognition. This paper mainly discusses the watershed algorithm. The over-segmentation occurrence is restrained through the dynamics combination rule. In addition, the segmentation testing has been conducted with the segmentation algorithm for blow holes and cracking defects, respectively, which, as a result, has achieved satisfactory segmentation effect. Finally, extension analysis is made for three-dimension engine data.Feature extraction is a key issue in pattern recognition. On the basis of defining the characteristic parameters for defect recognition and the quantitative calculation of main parameters, the welding defects are recognized with foregoing parameters.
|