Manufacturing deviation inspection method for ship block alignment structures based on terrestrial laser scanner data

Manufacturing deviation inspection for ship block alignment structures is a vital process in the construction of large-scale ships, the traditional inspection method is inefficient and labor-intensive. However, terrestrial laser scanning can create photorealistic 3D representations of any object, even in the most extreme outdoor conditions. This paper aims to inspect manufacturing deviation based on terrestrial laser scanner data. First, a plane segmentation process based on voxel clustering is applied, subdividing the point cloud into planes and extracts the region of interest. Then, point set registration is implemented and converted into probability density estimation problem of Gaussian mixture model, wherein point correspondences are assigned according to the posterior probability with mixed distance. Finally, parameter estimation is solved through Expectation-Maximum. The conducted experiments validate the higher efficiency of the proposed method compared to homogeneous algorithms and the feasibility of substituting the traditional method to accelerate the inspection process.