Research On Digital Twin Workshop Modeling Method Based On Point Cloud Dat

With the rapid development of the transformation and upgrading of the manufacturing industry,intelligent manufacturing has become an important trend in global manufacturing development.Building a digital workshop is one of the core aspects of achieving intelligent manufacturing.Digital twin workshop,as a typical model of digitalized workshops,still faces challenges in the modeling process such as large modeling workload,long cycle,and difficult scalability,resulting in low modeling efficiency and accuracy.Point cloud data has obvious advantages in the modeling field due to its concise mathematical representation and accurate object description ability,and has been widely concerned by scholars at home and abroad.Especially with the rapid development of deep learning,which provides powerful technical support for point cloud data analysis and processing.Therefore,this paper focuses on studying the method of digital twin workshop modeling based on point cloud data,aiming to achieve fast,efficient and accurate modeling of digital twin workshop.The main research contents are as follows:(1)After investigating the school’s training workshop,the height of the aircraft and the overlap degree of aerial photography required for the use of oblique photography data acquisition technology were determined.The DJI Phantom 4 RTK small multi-rotor highprecision unmanned aerial vehicle was used to collect point cloud scene data of the workshop.(2)Regarding the distribution characteristics of noise in the collected point cloud data of the workshop,the cloth simulation filter was first used to remove ground noise,and then the statistical filter was used to remove environmental noise.Finally,complete and separable workshop point cloud data was obtained.Based on this,a density-based clustering algorithm(Density-Based Spatial Clustering of Applications with Noise,DBSCAN)was used to segment the point cloud data of the workshop,taking into account the large differences in the density of the workshop point cloud data and the unknown number of clusters.The segmentation performance of the DBSCAN algorithm was explored under different parameter combinations,and ultimately,automatic and efficient segmentation of the point cloud data of the workshop was achieved.(3)To address the problem that the uneven number of sampled points in each machine tool point cloud data cannot meet the input requirements of deep learning algorithms,a method combining voxel-based downsampling and uniform downsampling is used to downsample the point cloud data,ensuring that the point cloud data meets the input requirements of the model while minimizing information loss.To address the issue of insufficient workshop machine tool point cloud data,data augmentation techniques such as point cloud rotation,translation,and random scaling are used to expand the machine tool point cloud dataset.As a result,a balanced machine tool point cloud dataset suitable for deep learning was obtained.(4)To address the problem that the small number of workshop machine tool point cloud data samples makes it difficult for deep learning algorithms to be effectively trained,an improved Point Net++ model that can effectively integrate local and global features is proposed and used as the shared network of a Siamese neural network to establish a smallsample learning model.To verify the effectiveness of the model,multiple sets of comparative experiments were conducted under different numbers of point cloud training samples.The experimental results show that the model not only effectively improves the classification accuracy on small datasets,but also has strong generalization ability.(5)Based on the research results of the preprocessing,segmentation,and classification of the workshop point cloud data mentioned above,a digital twin workshop modeling prototype system was developed.The system added functions such as location coordinate acquisition,automatic modeling,and model library management,which improved the prototype functions of the digital twin workshop modeling system.Ultimately,a low-cost,efficient,and automatic digital twin workshop modeling based on point cloud data was achieved.