With the development of deep learning, convolutional neural networks provide new solutions for obstacles in the industry. For seam weld grinding, traditional algorithms are neither convenient nor efficient. Meanwhile, previous methods are not robust for various shapes of the weld seam. In this paper, we propose a new algorithm based 1D convolutional neural network for 3D weld seam grinding. We test different loss functions for the 1D segmentation network and picked the best one for model training. Besides, we design various feature extracting blocks and make extensive experiments on the cloud point data set of the weld seam. The best combination of the loss function and feature extractor is generated for weld seam prediction. With the open operation on the 3D map and the elimination of abnormal points, we obtain a robust prediction grinding trail for the robot controller.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.