Abstract
Welding technology is a key manufacturing link in modern industrial steel structures, directly impacting welding quality and efficiency. This paper proposes LWSNet, a novel point cloud part segmentation network aimed at accurately segmenting weld areas across various joints to enhance automation and ensure quality. LWSNet builds upon the PointNet++ model, incorporating depthwise separable convolution and normalization-based attention module in its feature extraction block (FEB) for improved lightweight and feature extraction. A linear bottleneck structure and adaptive maximum pooling in its feature generation block (FGB) significantly reduce information loss during feature transmission. Additionally, a laser vision sensor model is designed for point cloud data acquisition, and the WeldJoint-PCD dataset is constructed using multiple data augmentation methods. Experimental results on WeldJoint-PCD show LWSNet’s superior segmentation effectiveness and lightweight design. On ShapeNetPart, cls.mIoU and ins.mIoU reach 82.9% and 85.5%, respectively, outperforming the original model by 1% and 0.4%, demonstrating LWSNet’s strong generalization capabilities.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
