基于稀疏表征的航空发动机铸件射线DR超分辨率重建算法研究

项目名称： 基于稀疏表征的航空发动机铸件射线DR超分辨率重建算法研究

项目编号： No.61261031

项目类型： 地区科学基金项目

立项/批准年度： 2013

项目学科： 无线电电子学、电信技术

项目作者： 邬冠华

作者单位： 南昌航空大学

项目金额： 38万元

中文摘要： 本项目在分析研究航空铸件DR检测图像退化原因的基础上，针对铸件图像细节对比度低，成像单元尺寸较大，造成裂纹类缺陷漏检的难题，采用基于多尺度字典的稀疏表示的方法，利用图像稀疏性的先验知识，使用已有的训练样本结合一定的构造算法分别构造出低分辨率图像和高分辨率图像样本的完备字典，利用稀疏表述算法求解待恢复图像在低分辨率图像字典上的稀疏表示系数，然后利用该稀疏表示系数结合高分辨率图像字典得到恢复后的高分辨率图像，达到有效检测裂纹类缺陷的目标。项目解决射线DR在铸件检测应用中存在的缺陷显示低分辨率和漏检的关键基础问题，瞄准航空铸件DR检测的需求，满足国家绿色制造技术的需要，应用对象明确，科学研究结果可以拓展应用到其他计算机成像的无损检测方法。

中文关键词： 射线DR；图像处理；图像稀疏算法；点扩散函数；图像恢复

英文摘要： The direct radiography(DR) image of aero engine casting with lower contrast in details and the units of image sensor maybe larger than defects were resulted in the defects such as cracks testing difficultly. The algorithm of super-resolution by spare represention will be studed in this item can be used to overcome the troubles in engine casting testing by DR. The algorithm based on the mode of image degradation caused by the scatter and other factors and a priori knowledge of image sparseness property which gained from image muti-scale spare represention dictionary.This dictionary will be maded by some image algorithm from the training samples with higher and lower resolution .The spare coefficients of higher and lower resolution image gained from spare represention algorithm. The spare coefficients combined with higher resolution image dictionary will be used to reconstruct super-resolution image,it can be achieve the high resolution image from low high resolution image, to attain super-resolution DR image which can be efficiently solved the trouble of cracks testing goal. This item will be overcome the difficulty of lower resolution in DR industrialization application.The study will be used to aero engine casting DR inspection and other non-destructive testing with computer image.

英文关键词： Direct Radiography；Image Processing；Image Sparse Algorithm；Point Spread Function；Image Restoration