基于改进PointNet++的船体分段合拢面构件智能识别算法研究

李瑞; 赵怡荣; 霍世霖; 汪骥; 史卫东

doi:10.19693/j.issn.1673-3185.03744

基于改进PointNet++的船体分段合拢面构件智能识别算法研究

Intelligent recognition algorithm for hull segment closure surface components based on improved PointNet++

摘要

摘要:
目的三维扫描仪获得的船体分段合拢面点云数据，具有精度高、数据量大的优势，能够很好的反映分段合拢面的建造状况。由于现有的PointNet++网络无法处理大容量点云数据，因此提出一种基于改进PointNet++的船体分段合拢面构件智能识别算法，实现针对大容量船体分段合拢面点云数据的构件智能识别。
方法基于超体素生长理论对船体分段合拢面点云数据分割及简化，构建船体分段合拢面点云数据集，并使用该数据集训练基于深度学习理论改进的PointNet++网络。
结果网络模型在船体分段合拢面点云数据训练集和测试集上的收敛结果趋于稳定，在测试集上准确率达到90%。
结论表明该方法具有良好的识别能力，能够完成船体分段合拢面构件智能识别。

Abstract:
Objectives The point cloud data of hull segment closure obtained by 3D scanner has the advantages of high precision and large data volume, and can reflect the construction status of segment closure well. Since the existing Pointnet++ network is unable to process large-capacity point cloud data, an intelligent recognition algorithm for hull segment closure surface components based on improved Pointnet++ is proposed to realize intelligent recognition of components for large-capacity hull segment convergence surface point cloud data.
Methods Based on hypervoxel growth theory, hull segment closure point cloud data is segmtioned and simplified, and hull segment closure point cloud data set is constructed, and the data set is used to train the Pointnet++ network improved based on deep learning theory.
Results The convergence results of the network model on the training and testing sets of point cloud data for hull segment closure surfaces tend to be stable, achieving an accuracy rate of 90% on the testing set.
Conclusions This indicates that the method has good recognition ability and can achieve intelligent recognition of hull segment closure surface components.

HTML全文

参考文献(11)

施引文献

资源附件(0)