基于灰度及形态特征融合的Micro-CT图像纤维束识别方法研究

Study on of fiber bundle recognition method of Micro-CT image based on fusion of gray-scale and morphological feature

利用Micro-CT成像以获得glass/epoxy(760E)编织复合材料的影像能够表征其织构形貌、内部缺陷,有助于研究复合材料微观结构与宏观力学性能之间的关系。但由于玻璃纤维与碳纤维在密度上与基体过于接近,导致Micro-CT图像中的灰度级差异过小且混乱,难以采用基于传统灰度特征的图像分割方法对纤维束所对应的图像区域进行识别。此外,由于制作工艺原因,正交两束纤维以及平行两束纤维过分挤压易造成工业CT图像中纤维图像区域相互混叠,加剧了对纤维束的准确识别的困难性。针对Micro-CT这种特殊的图像数据及纤维束区域分布相互混叠以及图像噪声的影响,采用多级图像处理及识别策略,将Micro-CT图像灰度特征及纤维束所表现出的形态特征进行融合,充分发挥灰度特征在噪声抑制及形态特征在纹理识别上的优势,并辅以形态学修正算法,实现了准确率较高的Micro-CT图像中纤维束识别。

The use of Micro-CT Imaging, for glass/epoxy (760E) braided composites image, has the ability to characterize their texture morphology and internal defects, and contributes to the research on the relations between the microstructure and macro-mechanical properties of composites. However , the glass fiber and carbon fiber are excessively close to the matrix in terms of density, resulting in smaller difference and confusing of grayscale in Micro-CT images. Hence, it is difficult to recognize the fiber bundle by using the grayscale features. Moreover, two excessively-squeezed orthotropic fibers and two parallel fibers could easily lead to mutual aliasing in industrial CT fiber images, making it difficult to correctly recognize the fiber bundle. Considering the characteristics of the Micro-CT images, this article utilizes multi-level image processing and recognition strategy to integrate the morphological features exhibited by grayscale features of fiber bundles from. The method we proposed which can take full advantage of the gray-scale features in noise suppression and the morphological features in texture recognition? and also can recognize the fiber bundle from Micro-CT images with a higher accuracy.