支持向量机在混凝土CT图像分析中的应用

Application of support vector machine to concrete CT image analysis

本文引入支持向量机的理论,对混凝土圆柱体试件一个扫描断面的5个应力阶段的CT图像进行了分析,对图像的各组成部分进行了分类,获得了清晰的裂纹、孔洞区域,砂浆区域和骨料区域的空间位置分布。从支持向量机分类图中可以看出裂纹或孔洞区域的大小分布随应力的变化过程,表明混凝土试件经历了压密、扩容、CT尺度裂纹扩展贯通至最后破坏的过程。与用传统的统计方法研究混凝土CT图像相比,支持向量机法不仅可以反映出材料的空间分布,同时使复杂的CT数分布对应转换为三类数的分布,简化了下一步统计分析。

The support vector machine is introduced to analyze the CT images of the concrete specimens with cylindrical cross section in 5 stages of deformation. The components of the CT images are classified and the spatial distributions of crack or cavity, motar and aggregate are obtained. The variation process of the relationship between distributions of crack or cavity magnitude and stress obtained from support vector machine classification maps clearly show that the specimens experience the process of condensed, volume expansion, crack propagation, coalescence and failure. The comparison of traditional statistical method for CT image analysis with the proposed method shows that the new method not only can reflect the spatial distribution of the materials but also converts the complicated CT data to corresponding distribution of 3 kinds of classification for simplifying the analysis.