Laser Scanning-Based Parameter Extraction for Ship Butt Weld Joints

Abstract To address the need for efficient and high-precision extraction of geometric parameters for three types of ship butt welds (I-groove, X-groove, and K-groove), an automatic weld parameter extraction algorithm based on laser scanning is proposed. First, point cloud surface separation is achieved through DBSCAN density clustering for I-groove welds and point cloud normal vector method for X-groove and K-groove welds. Subsequently, a universal parameter extraction module is constructed, incorporating a robust plate thickness calculation method and an adaptive groove angle extraction strategy. Finally, specialized parameter extraction strategies are designed according to the geometric characteristics of each weld type, forming a unified processing framework for multiple types of butt welds. Experimental results demonstrate that the mean absolute errors (MAE) for plate thickness and root gap of I-groove welds are 0.66 mm and 0.58 mm, respectively; the MAE for groove angles of X-groove and K-groove welds is maintained within 4°, while MAE for other parameters is controlled within 1 mm. Comparative experiments show that the proposed algorithm achieves superior measurement accuracy across all three weld types compared to traditional algorithms, with an 81.8% improvement over the slope analysis method, a 73.9% improvement over the height transition feature method, and an overall average accuracy improvement of approximately 77.9%.