【作者】 李俊；

【导师】 严海蓉；

【作者基本信息】 北京工业大学 ， 软件工程（专业学位）， 2022， 硕士

【摘要】 逆向制造是基于样品批量生产产品的过程。它需要对现有样品进行扫描或者人工测量,再画图获得生产图纸,最后加工制造。人工画图工作量大,时间周期比较长。然而大多数的产品设计是对已有成果的借鉴和重用。设计者从现有图纸得到启发,或者在现有产品图纸上进行修改,可避免大量重复性设计工作。当前,大部分的工程图纸推荐都是以文本为基础的,存在文本不能完全表达工程图纸具体内容的问题。针对上述问题,本研究设计并实现产品设计推荐软件工具系统。通过三维点云扫描仪采集样品点云数据,在点云数据的基础上进行工程图纸的检索推荐。本研究可以缩短产品设计开发周期、改善产品质量并降低成本。本文的研究工作分为以下几个方面的内容:(1)设计并实现一套可视化系统。使用Qt图形框架和视觉工具函式库VTK进行软件图形界面和产品点云数据的可视化,使设计人员能清楚地观察点云数据和工程图纸数据,并且能以自由旋转的视角观察三维模型,选择投影视角。(2)研究产品点云数据处理。使用三维扫描仪扫描企业产品,采集产品点云数据。研究点云滤波,包括直通滤波、离群点滤波、体素滤波及半径滤波等点云滤波方法。研究点云曲面重建,生成产品点云数据曲面模型,得到产品点云数据的多视图。(3)提出基于产品点云数据和视觉词袋模型的工程图纸推荐方法。建立工程图纸数据集,采用SIFT算法从不同种类的工程图纸提取视觉词汇特征。使用K-Means聚类算法对提取的SIFT特征聚类。构造视觉词汇向量,建立视觉词袋词典。构建工程图纸的视觉词袋模型,实现工程图纸的推荐。使用产品点云数据的多视图,提取产品视图轮廓特征,检索相似工程图纸,推荐给设计者。本研究基于逆向制造,设计并实现产品设计推荐软件工具。系统可以代替现有点云数据处理和工程图纸推荐软件,解决从扫描的点云数据到工程图纸设计推荐的过程。更多还原

【Abstract】 Reverse manufacturing is the process of mass-producing products based on samples.It requires scanning or manual measurement of existing samples,then drawing to obtain production drawings,and finally processing and manufacturing.Manual drawing has a large workload and a long time cycle.However,most new designs are based on the reference and reuse of existing design knowledge.Designers take inspiration from existing drawings,or modify existing product drawings,avoiding a lot of repetitive design work.At present,most of the engineering drawing recommendations are based on text,and there is a problem that the text cannot fully express the specific content of the engineering drawings.Aiming at the problems,this research designs and implements a product design recommendation software tool system.The sample point cloud data is obtained through a 3D point cloud scanner,and the engineering drawings are retrieved and recommended on the basis of the point cloud data.This research shortens product design and development cycles,improves quality of products and reduces costs.The main research work is divided into the following aspects:Firstly,this research designs and implements a visualization system.This research uses the Qt and VTK to visualize the software graphical interface and product point cloud data,so that designers can clearly observe the point cloud data,engineering drawing data and 3D model from a freely rotating perspective,and choose a projection perspective.Secondly,this thesis researches point cloud data of product processing.a 3D scanner is used to scan enterprise products and collect product point cloud data.This thesis researches point cloud filtering,including straight-through filtering,outlier filtering,voxel filtering,radius filtering and other filtering methods.This thesis researches the surface reconstruction of point cloud,generates surface model of point cloud data,and obtains multi-view of point cloud data.Thirdly,this thesis proposes a recommendation method for engineering drawing retrieval based on the bag of visual words model and point cloud data of product.A dataset of engineering drawings is established,and the visual vocabulary features are extracted from different kinds of engineering drawings by SIFT algorithm.The extracted SIFT features are clustered by K-Means clustering algorithm.A bag of visual words model of engineering drawings is constructed to realize the recommendation of engineering drawings.This thesis uses the multi-view of product point cloud data to extract the contour features of product views,retrieve similar engineering drawings,and recommend them to designers.Based on reverse manufacturing,this research designs and implements a product design recommendation software tool.The system can replace the existing point cloud data processing and engineering drawing recommendation software,and solve the process from scanned point cloud data to engineering drawing design recommendation.更多还原