炭制品图像的缺陷分割与样本提取

Defect Segmentation and Specimen Extraction of Carbon Product Image

针对炭制品X光图像的特点,为快速准确地分割出缺陷,提出了基于迭代的阈值构造方法和数学形态学相结合的缺陷分割方法.通过迭代算法确定了图像分割的最佳阈值,有效地减少了噪声对分割阈值的影响;设计可疑点区域搜索算法,确定了可疑点区域的位置和大小;通过试验得到了滤波模板的最佳概率阈值,并建立了可疑点的合并条件表达式,得到了完整的缺陷区域.研究结果表明:该法很好地解决了噪声抑制和保持图像缺陷细节之间的矛盾,且缺陷分割结果非常准确.最后,在缺陷分割的基础上,研究了缺陷样本的输出方式,设计了两个三维特征矩阵存储缺陷特征信息,实现了缺陷几何特征与灰度特征的提取,为进一步的缺陷特征参数的提取与缺陷识别奠定了基础.

In order to segmentation defects quickly and exactly, according to the characteristic of x-ray detection images of carbon product, defect segmentation method that threshold-construction based on iteration links mathematics morphology is advanced. Optimal threshold of image segmentation is ensured by iteration algorithm, the effect of noise is decreased greatly. The position and size of suspicious area are made certain using search algorithm, and the suspicious area is marked by minimal rectangle. The optimal probability threshold of filter mask is gotten by mean of experiment, the full defect area is gotten because of the merged expression of suspicious points. The results demonstrate that the method could suppress noise as much as possible while preserving fine details, and the result of defect segmentation is fine. Finally, based on defect segmentation the output mode of defects specimen is studied, two three dimensional matrixes that deposit characteristic information of defects are designed, geometry and grey characteristic are extracted successfully, which will lay a good foundation for flaw feature parameter extraction and defect recognition. 14figs., 15refs.