基于定性视觉特征的裂纹缺陷的仿真

Simulation of flaw defect based on qualitative vision features

提出一种基于定性视觉特征的裂纹缺陷X光图像的仿真方法。从裂纹的定性视觉特征提取语义,进行逻辑描述,生成裂纹缺陷的骨架,对骨架各个像素分别进行运动模糊滤波,由所需裂纹的大致宽度对模糊骨架进行偏移后加权叠加。根据裂纹缺陷整体走向对整幅图进行运动模糊滤波,引入模糊几何特征,以调整仿真算法参数。实验结果表明,该方法可得到具备模糊性和随机性的各类裂纹缺陷。

A new approach for simulation of flaw defect X-ray images based on qualitative vision features is presented. Firstly, semantic extraction from the qualitative vision features and logical description are proposed. Secondly, the skeleton of flaw defect is generated. Thirdly, motion filter is applied to each pixel on skeleton. Then, according to the rough flaw width needed, fuzzy skeletons are weighted stacked after migration. Finally, the whole image is filtered considering the integral direction. Fuzzy geometric features are introduced to adjust the parameters of simulation algorithm. The experimental results show that the approach can produce a large range of simulated flaw defects with randomness and fuzziness.