Research Of Turbine Blade Defects Detection Based On 3D ICT Images

Hollow turbine blade nondestructive testing and analysis is the key technology for breaking through the bottleneck of developing the new generation aero blade and assuring its quality. Due to the complex inner structure of turbine blades, the traditional defect detection methods can hardly satisfy the inspection need, and the detection technique based on the ICT has become a hot topic in the field of blade testing. According to the progress made in the area of image segmentation, image sequence processing and scientific visualization, the thesis investigates turbine blade defects detection technologies based on 3D ICT image sequences and the corresponding software prototype system is developed. The main research content and contribution include the following aspects:1. For the serious noise in practical ICT images, a preprocessing step is adopted for slice images. After analyzing source and characteristic of the noise in ICT images, the thesis establishes the mathematical models for noise simulation. As to the blurring and low contrast of defects target, a hybrid filter by estimating Gaussian noise variance is proposed for image sequences. By transferring the Gaussian noise variance parameter between layers and employing different filters for different sort of pixels, the algorithm can effectively improve the precision and accuracy, as well as preserving edge details.2. Aiming at defects detection for low-contrast ICT images, a comprehensive topology-region based image segmentation algorithm is presented by combining multi-threshold segmentation, connected region labeling and region growing methods. This method converts the low contrast of the whole image into a higher contrast in different topological structures, and improves the precision and accuracy of segmentation. By analyzing gray distribution of low contrast defect targets and noise objectives, a templates segmentation algorithm is proposed for images with serious noises. This method improves defect detection accuracy and can distinguish noise objectives and defects target.3. For the ambiguity existing in the 2D defect detection, a method of defects detection for 3D ICT images is proposed. According to the isotropy of ICT images and the small shape change of defect entities between image layers, this algorithm...