逆向工程中点云修补与曲面反求的研究
|
摘要
逆向工程(Reverse Engineering,RE)也称反求工程,是消化、吸收先进技术时采取的一系列分析方法和应用技术的结合。作为新产品开发和消化、吸收先进技术的重要手段,逆向工程(RE)的研究正受到各国工业和学术界的高度重视。并已经在汽车,模具,家电,文物修复与仿制等领域得到了广泛的运用。
逆向工程中的产品建模一般按照如下流程进行:通过测量使实物成为数字化的点云,测量数据点云的预处理,CAD模型重构。在上述这些步骤中,点云质量的好坏和曲面拟合重构方案的选择成为最终影响CAD模型质量的关键性因素。
在用激光三维扫描仪对被扫描对象扫描的过程中,由于受到扫描对象、现场环境、扫描速度和时间等的限制,导致三维物体的某些部分获取的数据不完整,造成三维数据的残缺。本文将探讨两种测量数据的修补方法:基于截面数据的测量数据修补方法和基于三维散乱数据的修补方法。基于截面测量数据的修补方法通过把某一坐标轴方向坐标等值化处理,将三维数据转化为基于神经网络二维曲线拟合的方式;另外一种基于BP网络的三维散乱点云的神经网络修补方法,采用将三维曲面某一坐标看成是其他两个坐标的输出的方式直接对点云坐标进行预测。
另外,本文将针对现有的工程实际中使用的曲面拟合方案中不适应处理数据量庞大的点云或与现有CAD软件的兼容性问题,探讨了一种曲面拟合流程,采用将曲线、曲面在不同环境下分步拟合的策略。从而综合利用了专业反求软件点云处理能力强、曲线拟合质量高和大型CAD软件面拟合手段丰富、建立体模型时兼容性好的特点。
研究表明,基于截面数据的测量数据修补方法和基于三维散乱数据的修补方法取得了较好的效果,丰富了不同形态点云的修补手段;新的曲面拟合的方案较大程度上克服了传统方案的弊端,探索了一种具有实际工程运用价值的逆向流程。
Reverse Engineering (Reverse Engineering RE), also known as reverse engineering, considered as a feedback-loop design from upstream to downstream, and then develop more advanced technology products, as well as absorb advanced technologies from series of analytical methods and application technology.
As an important method of developing new products and absorbing advanced technologies. Reverse-engineering (RE) is understudied by industrial field and academia in many countries. So far, it have applied in wide range of fields such as vehicles, mold , household appliances , the restoration and imitation of cultural heritage. Industrial engineers, in the use of free-form surface reconstruction technique, can significantly shorten the product design and manufacturing cycles, improving product development efficiency.
Modeling of the products in Reverse Engineering is in accordance with the following general procedures: acquire the data-point cloud through measuring the product, Pretreatment data-point, CAD model reconstruction. In these steps, The quality of point cloud and surface fitting reconstruction program eventually to be a crucial factor affecting the quality of CAD models.
During the scanning process by 3D laser scanner, due to the limit of scan target site, scanning speed and time , data from certain parts of the 3D object was not in integrity, which caused defects of 3D data. Some scholars have already given a cross-section measurements remedy , used to coordinate the equivalent of a particular axis direction, in this way,3D data into 2D curve fitting based on neural network approach. But its application is very narrow, Because the nature of the cloud patterns must followed to a stringently limit. This article will explore another unorganized point cloud repairing method based on BP neural network, and compared these two methods. At last ,give an introduction for the methods on will be of various types of cloud shows targeted for repairing incomplete data.
In addition, the surface fitting programs in actual engineering can divided into two major categories :one is directly rebuild the surface base on professional software, the other is using CAD software to Complete the surface-reconstruction based on data point cloud, such as Pro-E. I-DEAS and UGII. These existed programs are not suited to deal with big amount of data point or have compatibility problems with CAD soft wares. This article will explore a new surface fitting program, taken into account of two items: professional reverse software‘s processing ability on data point cloud and CAD software’s large high-quality surface fitting means, which also have good compatibility features with other CAM soft wares. For Specific, the fitting of the curve and the surface are under different environments. This strategy resolve these contradictions. and the merits of the two programs are integrated. This article gives a rather good treatment process, and also have actual value in implementation.
Research shows that: the three-dimensional artificial neural network repairing on data-point cloud achieved good results, enriched the forms of repairing ways; The new surface fitting program greatly overcome the shortcomings of the traditional program, give a valuable exploration which have actual implementation value in engineering.
逆向工程中的产品建模一般按照如下流程进行:通过测量使实物成为数字化的点云,测量数据点云的预处理,CAD模型重构。在上述这些步骤中,点云质量的好坏和曲面拟合重构方案的选择成为最终影响CAD模型质量的关键性因素。
在用激光三维扫描仪对被扫描对象扫描的过程中,由于受到扫描对象、现场环境、扫描速度和时间等的限制,导致三维物体的某些部分获取的数据不完整,造成三维数据的残缺。本文将探讨两种测量数据的修补方法:基于截面数据的测量数据修补方法和基于三维散乱数据的修补方法。基于截面测量数据的修补方法通过把某一坐标轴方向坐标等值化处理,将三维数据转化为基于神经网络二维曲线拟合的方式;另外一种基于BP网络的三维散乱点云的神经网络修补方法,采用将三维曲面某一坐标看成是其他两个坐标的输出的方式直接对点云坐标进行预测。
另外,本文将针对现有的工程实际中使用的曲面拟合方案中不适应处理数据量庞大的点云或与现有CAD软件的兼容性问题,探讨了一种曲面拟合流程,采用将曲线、曲面在不同环境下分步拟合的策略。从而综合利用了专业反求软件点云处理能力强、曲线拟合质量高和大型CAD软件面拟合手段丰富、建立体模型时兼容性好的特点。
研究表明,基于截面数据的测量数据修补方法和基于三维散乱数据的修补方法取得了较好的效果,丰富了不同形态点云的修补手段;新的曲面拟合的方案较大程度上克服了传统方案的弊端,探索了一种具有实际工程运用价值的逆向流程。
Reverse Engineering (Reverse Engineering RE), also known as reverse engineering, considered as a feedback-loop design from upstream to downstream, and then develop more advanced technology products, as well as absorb advanced technologies from series of analytical methods and application technology.
As an important method of developing new products and absorbing advanced technologies. Reverse-engineering (RE) is understudied by industrial field and academia in many countries. So far, it have applied in wide range of fields such as vehicles, mold , household appliances , the restoration and imitation of cultural heritage. Industrial engineers, in the use of free-form surface reconstruction technique, can significantly shorten the product design and manufacturing cycles, improving product development efficiency.
Modeling of the products in Reverse Engineering is in accordance with the following general procedures: acquire the data-point cloud through measuring the product, Pretreatment data-point, CAD model reconstruction. In these steps, The quality of point cloud and surface fitting reconstruction program eventually to be a crucial factor affecting the quality of CAD models.
During the scanning process by 3D laser scanner, due to the limit of scan target site, scanning speed and time , data from certain parts of the 3D object was not in integrity, which caused defects of 3D data. Some scholars have already given a cross-section measurements remedy , used to coordinate the equivalent of a particular axis direction, in this way,3D data into 2D curve fitting based on neural network approach. But its application is very narrow, Because the nature of the cloud patterns must followed to a stringently limit. This article will explore another unorganized point cloud repairing method based on BP neural network, and compared these two methods. At last ,give an introduction for the methods on will be of various types of cloud shows targeted for repairing incomplete data.
In addition, the surface fitting programs in actual engineering can divided into two major categories :one is directly rebuild the surface base on professional software, the other is using CAD software to Complete the surface-reconstruction based on data point cloud, such as Pro-E. I-DEAS and UGII. These existed programs are not suited to deal with big amount of data point or have compatibility problems with CAD soft wares. This article will explore a new surface fitting program, taken into account of two items: professional reverse software‘s processing ability on data point cloud and CAD software’s large high-quality surface fitting means, which also have good compatibility features with other CAM soft wares. For Specific, the fitting of the curve and the surface are under different environments. This strategy resolve these contradictions. and the merits of the two programs are integrated. This article gives a rather good treatment process, and also have actual value in implementation.
Research shows that: the three-dimensional artificial neural network repairing on data-point cloud achieved good results, enriched the forms of repairing ways; The new surface fitting program greatly overcome the shortcomings of the traditional program, give a valuable exploration which have actual implementation value in engineering.
引文
[1]田晓东,史桂蓉,阮雪榆.复杂曲面实物的逆向工程及其关键技术.机械设计与制造工程.2000,(4) .1～6
[2]戴德山.RE在产品快速成形中的应用.机电产品开发与创新.2003,(5).33～34
[3]马正源,杜丽明.基于逆向工程的自由曲面建模.计算机应用与技术.2004,(3) .55～58
[4]罗里荣.面向快速原型制造的形状反求关键技术的研究:[硕士学位论文].秦皇岛:燕山大学,2004.
[5] Chivate, P N, Jablokow, A G.Solid model generation from measured point data.Computer Aided Design.2000,(21).587-600
[6]刘之升,黄纯.反求工程技术.第1版.北京:机械工业出版社,1992.59～62
[7]雷刚,邹昌平,黄剑鸣.快速逆向工程研究进展.机械工艺师.2000,(3) .84～86
[8]吴福生,丁玉成,卢秉恒.基于点云数据复杂曲面产品的快速开发.电加工与模具.2005,(4).32～35
[9]朱东波,张舜德,李涤尘等.密集散乱测量数据点的B样条曲面拟合研究.计算机辅助设计与图形学学报.2001,13(12).1123～1128
[10] Edgeworth R, Wilhelm R., Adaptive sampling for coordinate metrology. Precision Engineering 1999, ( 23).144～54
[11] Hsi-Yung Feng, Yixin Liu, Fengfeng Xi, Analysis of digitizing errors of a laser scanning system. Precision Engineering 2001, (25) .185~191.
[12]施法中.计算机辅助几何设计与非均匀有理B样条.第2版.北京:北京航空航天大学出版社,2001.20~45
[13] Zexiao, Xie; Chengguo, Zhang; Qiumei, Zhang, A simplified method for the extrinsic calibration of structured-light sensors using a single-ball target. International Journal of Machine Tools and Manufacture (2004), 44(11): 1197-1203.
[14] Diaa F. ElKott, Stephen C. Veldhuis, Isoparametric line sampling for the inspection planning of sculptured surfaces. CAD 2005, (37) .189～200
[15] Edgeworth R, Wilhelm R., Adaptive sampling for coordinate metrology. Precision Engineering.1999,(23).144～154
[16] Curless B,Levoy M. A volumetric method for building complex models from range images [A]. SIGGRAPH’96 Proceedings [C]. New Orleans, USA: 1996, 303～312
[17]金涛,陈建良,童水光.逆向工程技术研究进展.中国机械工程.2002,13(16).1430～1436
[18]李东泽.反求工程曲面重构技术研究:[硕士学位论文].大连:大连理工大学,2002.
[19]刘晨光.基于神经网络的自由曲面重建:[硕士学位论文].保定:华北电力大学硕士论文,2001.
[20]马正元,蔡鹤皋.三维数字化测量系统.中国机械工程.2000,11(10) .1161～1164
[21]冯涛,吴晓林,王亚等.反求工程与快速成型.机械工艺师.1998,(10).4～6
[22] Amenta N,Bern M, Kamvysselis M.A new Voronoi-based surface reconstruction algorithm [A] .SIGGRAPH’98 Proceedings [C].New York, USA:1998.415～421
[23]孙玉文.面向快速原型制造的反求工程关键技术研究:[博士学位论文].大连:大连理工大学,2001.
[24]栗全庆,王宏,张英杰等.实物反求工程的关键技术分析.机械设计.1999,(6).4～6
[25]周儒荣,张丽艳,苏旭等.海量散乱点的曲面重建算法研究.软件学报.2001,12(2) .249～255
[26] Carr J, Beatson R, Cherrie J, et al.Reconstruction and representation of 3D objects with radial basis functions [A].SIGGRAPH’01 Proceedings [C].Los Angeles, USA: 2001.67～76
[27] C.Kahn, P.Porras, SC.Staniford.Comman Intrusion Detection Framework.Journal of Computer Security.1998,(7).242~267
[28] Rumelhart.D.E, Hinton.G.E,Williams.R.J,Learning.Representation by Back propagation Errors.Nature.1986, 6188(323).533~523
[29] Milroy M J, Bradley C, Vickers G W.Segmentation of Wrap-around Model Using an Active Contour.Computer Aided Design.1997,29(4).299～320
[30]许东,吴铮.基于MATLAB的神经网络系统分析与设计.第1版.西安:西安电子科技大学出版社,2000.50～80
[31]陈怀琛.MATLAB在理工科课程中的应用指南.第2版.西安:西安电子科技大学出版社,2003.80～105
[32] Vida J, Martin RP,Varady T.A surface Blending Methods that Use Parametric Surface.Computer Aided Design.1994,26(5).287~298
[33] Meyer M, Debunne G, Desbrun M, et al.Fast Animation and Control of No rigid Structures.Computer Graphics.1990, 24 (4).243~252
[34]金涛,单岩,童水光.产品反求工程中基于几何特征及约束的模型重建.计算机辅助设计与图形学学报.2001,13(3).202～217
[35] WangJ,Oliveira M M.A hole-filling strategy for reconstruction of smooth surfaces in range images [A].In: Proceedings of Computer Graphics and Image Processing [C].S o Carlos, Brazil: 2003.11～18
[36] LIN CY, LIOU C S,LAI J Y.Surface-lofting Approach for Smooth- surface reconstruction From 3D Measurement a Computer In Industry.1997,34(1).73-85
[37]黄小平,杜晓明,熊有伦.逆向工程中的建模技术.中国机械工程.2001,12(5) .539~542
[38]王文深,狄文辉.反求工程技术及其应用.河南机电高等专科学校学报.2001,9(2) .6~9
[39]张磊,胡春,钱锋.BP算法局部极小问题改进的研究发展工业控制计算机,2004, 17(9).33~34
[40]吕俊,张兴华.几种快速BP算法的比较研究.现代电子技术.2003,24(9).36~42
[41]下赞松,许洪国.快速收敛的BP神经网络算法.吉林大学学报.2003,33(4): 80~86
[42]刑渊.集成反向工程系统及其关键技术的研究:[博士学位论文].上海:上海交通大学,1998.
[43] SHI Guirong, XING Yuan.Reverse-Engineering Automatic Geometry Modeling System.Journal Shanghai Jiaotong University.2000,34(3).392~395
[44]孙印节,田效伍.Pro/Engineer基础与实例教程.第1版.北京:电子工业出版社,2004.76~92
[45]李江雄.复杂曲面产品反求工程的CAD建模技术:[博士学位论文].杭州:浙江大学,1998.
[46]杨民,杨海成.实物测量造型技术中的数据分块方法.机械科学与技术.1996,15(6).989~992
[2]戴德山.RE在产品快速成形中的应用.机电产品开发与创新.2003,(5).33～34
[3]马正源,杜丽明.基于逆向工程的自由曲面建模.计算机应用与技术.2004,(3) .55～58
[4]罗里荣.面向快速原型制造的形状反求关键技术的研究:[硕士学位论文].秦皇岛:燕山大学,2004.
[5] Chivate, P N, Jablokow, A G.Solid model generation from measured point data.Computer Aided Design.2000,(21).587-600
[6]刘之升,黄纯.反求工程技术.第1版.北京:机械工业出版社,1992.59～62
[7]雷刚,邹昌平,黄剑鸣.快速逆向工程研究进展.机械工艺师.2000,(3) .84～86
[8]吴福生,丁玉成,卢秉恒.基于点云数据复杂曲面产品的快速开发.电加工与模具.2005,(4).32～35
[9]朱东波,张舜德,李涤尘等.密集散乱测量数据点的B样条曲面拟合研究.计算机辅助设计与图形学学报.2001,13(12).1123～1128
[10] Edgeworth R, Wilhelm R., Adaptive sampling for coordinate metrology. Precision Engineering 1999, ( 23).144～54
[11] Hsi-Yung Feng, Yixin Liu, Fengfeng Xi, Analysis of digitizing errors of a laser scanning system. Precision Engineering 2001, (25) .185~191.
[12]施法中.计算机辅助几何设计与非均匀有理B样条.第2版.北京:北京航空航天大学出版社,2001.20~45
[13] Zexiao, Xie; Chengguo, Zhang; Qiumei, Zhang, A simplified method for the extrinsic calibration of structured-light sensors using a single-ball target. International Journal of Machine Tools and Manufacture (2004), 44(11): 1197-1203.
[14] Diaa F. ElKott, Stephen C. Veldhuis, Isoparametric line sampling for the inspection planning of sculptured surfaces. CAD 2005, (37) .189～200
[15] Edgeworth R, Wilhelm R., Adaptive sampling for coordinate metrology. Precision Engineering.1999,(23).144～154
[16] Curless B,Levoy M. A volumetric method for building complex models from range images [A]. SIGGRAPH’96 Proceedings [C]. New Orleans, USA: 1996, 303～312
[17]金涛,陈建良,童水光.逆向工程技术研究进展.中国机械工程.2002,13(16).1430～1436
[18]李东泽.反求工程曲面重构技术研究:[硕士学位论文].大连:大连理工大学,2002.
[19]刘晨光.基于神经网络的自由曲面重建:[硕士学位论文].保定:华北电力大学硕士论文,2001.
[20]马正元,蔡鹤皋.三维数字化测量系统.中国机械工程.2000,11(10) .1161～1164
[21]冯涛,吴晓林,王亚等.反求工程与快速成型.机械工艺师.1998,(10).4～6
[22] Amenta N,Bern M, Kamvysselis M.A new Voronoi-based surface reconstruction algorithm [A] .SIGGRAPH’98 Proceedings [C].New York, USA:1998.415～421
[23]孙玉文.面向快速原型制造的反求工程关键技术研究:[博士学位论文].大连:大连理工大学,2001.
[24]栗全庆,王宏,张英杰等.实物反求工程的关键技术分析.机械设计.1999,(6).4～6
[25]周儒荣,张丽艳,苏旭等.海量散乱点的曲面重建算法研究.软件学报.2001,12(2) .249～255
[26] Carr J, Beatson R, Cherrie J, et al.Reconstruction and representation of 3D objects with radial basis functions [A].SIGGRAPH’01 Proceedings [C].Los Angeles, USA: 2001.67～76
[27] C.Kahn, P.Porras, SC.Staniford.Comman Intrusion Detection Framework.Journal of Computer Security.1998,(7).242~267
[28] Rumelhart.D.E, Hinton.G.E,Williams.R.J,Learning.Representation by Back propagation Errors.Nature.1986, 6188(323).533~523
[29] Milroy M J, Bradley C, Vickers G W.Segmentation of Wrap-around Model Using an Active Contour.Computer Aided Design.1997,29(4).299～320
[30]许东,吴铮.基于MATLAB的神经网络系统分析与设计.第1版.西安:西安电子科技大学出版社,2000.50～80
[31]陈怀琛.MATLAB在理工科课程中的应用指南.第2版.西安:西安电子科技大学出版社,2003.80～105
[32] Vida J, Martin RP,Varady T.A surface Blending Methods that Use Parametric Surface.Computer Aided Design.1994,26(5).287~298
[33] Meyer M, Debunne G, Desbrun M, et al.Fast Animation and Control of No rigid Structures.Computer Graphics.1990, 24 (4).243~252
[34]金涛,单岩,童水光.产品反求工程中基于几何特征及约束的模型重建.计算机辅助设计与图形学学报.2001,13(3).202～217
[35] WangJ,Oliveira M M.A hole-filling strategy for reconstruction of smooth surfaces in range images [A].In: Proceedings of Computer Graphics and Image Processing [C].S o Carlos, Brazil: 2003.11～18
[36] LIN CY, LIOU C S,LAI J Y.Surface-lofting Approach for Smooth- surface reconstruction From 3D Measurement a Computer In Industry.1997,34(1).73-85
[37]黄小平,杜晓明,熊有伦.逆向工程中的建模技术.中国机械工程.2001,12(5) .539~542
[38]王文深,狄文辉.反求工程技术及其应用.河南机电高等专科学校学报.2001,9(2) .6~9
[39]张磊,胡春,钱锋.BP算法局部极小问题改进的研究发展工业控制计算机,2004, 17(9).33~34
[40]吕俊,张兴华.几种快速BP算法的比较研究.现代电子技术.2003,24(9).36~42
[41]下赞松,许洪国.快速收敛的BP神经网络算法.吉林大学学报.2003,33(4): 80~86
[42]刑渊.集成反向工程系统及其关键技术的研究:[博士学位论文].上海:上海交通大学,1998.
[43] SHI Guirong, XING Yuan.Reverse-Engineering Automatic Geometry Modeling System.Journal Shanghai Jiaotong University.2000,34(3).392~395
[44]孙印节,田效伍.Pro/Engineer基础与实例教程.第1版.北京:电子工业出版社,2004.76~92
[45]李江雄.复杂曲面产品反求工程的CAD建模技术:[博士学位论文].杭州:浙江大学,1998.
[46]杨民,杨海成.实物测量造型技术中的数据分块方法.机械科学与技术.1996,15(6).989~992