机构地区: 华南理工大学电子与信息学院
出 处: 《现代电子技术》 2011年第2期79-81,86,共4页
摘 要: 提出一种基于定性视觉特征的疏松型缺陷X光图像的生成方法。算法采用基于定性视觉特征的原理,先选定缺陷仿真的区域,设计缺陷模板的大小;然后提取部分高斯噪声点,确定缺陷位置及中心亮点位置,以及进行低通滤波和几何扩散,生成独立于背景的仿真缺陷;将此仿真缺陷叠加到铸件图像上与背景融合,完成了实际铸件缺陷仿真。实验结果表明,所生成缺陷在形状和灰度方面都具有随机性,且逼近真实缺陷。 A new approach for generation of casting porosities defect X-ray images based on qualitative vision features is presented. The algorithm is based on qualitative vision features. The generation process of the image is that the simulation re- gion of defects is selected, the size of the defect template is designed, some Gauss noice points are extracted, the positons of every small flaw and its center can be determined, the simulation defects are produced by low-pass tilting and geometric sprea- ding, and then the simulation flaws are overlapped to the real casting image to complete the simulation. The experimental resuits show that the generated images have their randomness in shapes and grey levels, and approximate the real defects.