为了获得高精度工业CT尺寸测量结果，针对工业结构中常见的圆柱体提出了一种亚体素CT图像尺寸测量方法。首先使用MarchingCubes方法检测边缘点和法线，然后顶点沿着法线方向移动到梯度最大值处，最后采用点云分割方法获得圆柱体直径。在仿真数据中验证了该方法的测量精度优于全局表面测定方法的测量精度，具有一定的工程应用价值。

To obtain high-accuracy dimensional metrology results of industrial CT, a dimensional metrology method of sub-voxel CT image was proposed for cylinder, which was a common geometric structure in the industry. Firstly, the MarchingCubes method was used to detect the edge points and normal lines. Then the vertices were moved along the normal direction to the maximum gradient. Finally, the point cloud segmentation method was used to get the diameter of the cylinder. The measurement accuracy of this method had been verified in simulation data to be superior to that of the global surface measurement method, and which had a certain engineering application value.