Optimisation of continuous overlapping point cloud data for laser cladding

To address the issue where 3D scanning of complex structural metallic targets during the laser cladding process frequently generates overlapping noisy point cloud data that significantly deviates from the inherent characteristics of the object, this paper proposes a multi-layer point cloud extraction and data optimization method to eliminate overlapping noise and restore the original scanning data. The proposed method is based on point cloud surfaces extracted using the Random Sample Consensus (RANSAC) algorithm. By delineating object boundaries, the point cloud surface is segmented into distinct point cloud blocks. Subsequently, noisy blocks are identified based on their overlapping nature and limited size, and these blocks are systematically removed in a block-by-block manner. As the method exclusively processes noisy point cloud blocks, it preserves the integrity, precision, and original morphology of the non-noisy point cloud data.