This paper introduces a Joint Bilateral Guided filtering algorithm with Dynamic Neighborhood Selection(JBG-DNS) for noise reduction in point cloud data. The proposed algorithm intelligently selects points within a neighborhood and incorporates the principles of bilateral filtering, effectively eliminating noise while preserving detailed features. This provides high-quality data for subsequent applications such as weld seam measurement. Experimental results demonstrate the algorithm’s outstanding performance in point cloud denoising, significantly enhancing the accuracy of weld seam measurement. Compared to traditional manual measurement, non-contact weld seam measurement utilizing point cloud data offers advantages of non-invasiveness, efficiency, and automation, providing robust support for welding process monitoring and quality control.