Industrial CT linear array scanning is an important method for acquiring the internal characteristic structures of aero-engine turbine blades, and extracting the contours of reconstructed tomographic grayscale images is a key step for measuring dimensions such as blade wall thickness. Since the commonly used pixel-level unsupervised evaluation methods suffer from blurred extracted edges and low dimensional measurement accuracy, this paper proposes a subpixellevel contour extraction algorithm based on intelligent parameter optimization for CAD model matching. Firstly, the local binary fitting (LBF) geometric active contour model is employed to extract edges; Secondly, the corresponding crosssectional point cloud is acquired from the CAD model; Thirdly, the coordinates of the two are unified using the oriented bounding box (OBB) algorithm; And finally, the evaluation function is constructed based on the Hausdorff distance. Ultimately, four parameters in the LBF model are optimized via the dung beetle optimizer (DBO), thereby achieving optimal contour extraction. The results of CT tomography images of turbine blades show that the relative error is less than 1.6%, compared with traditional edge detection algorithms such as Canny, Ostu, and Zernike, the method proposed in this paper can significantly improve the measurement accuracy.