视觉检测技术在宝石毛坯三维测量中的应用研究
|
摘要
机器视觉的重要任务之一就是设法在一般的二维灰度图像中采用距离和深度测量方法求得第三维坐标(通常是用Z 坐标表示) , 从而形成了三维视觉。由于实现思想和条件的不同, 就产生了相应的诸多视觉方法【37】-【42】。其中, 结构光三维视觉以其大量程、大视场、较高精度、光条图像信息提取简单、实时性强及主动受控性好等特点, 近年来在工业视觉检测中得到了愈来愈广泛的应用。
本文介绍了结构光三维视觉原理, 并在广义坐标系中采用透视投影变换理论建立了简单合理的基于线结构光的三维测量数学模型,并根据宝石的物理特性设计了线结构光三维视觉测量系统。最后用一个高精度的已知其半径与高度的轴承圆柱滚动体作为实验物进行三维测量,测量的数据走向与条纹的形状是符合的,因此说整个系统的设计理念是正确的。
structured-light method is to transform the light beam that the laser instrument is sent out into different geometry shape light through the optical system, including point, mongline, multithreading,single round , concentric more round,gridding cross etc. casts to scenery, forms the pattern and ingested by the camera on the scenery (this method is called initiative sensing method), and then calculates , gets the depth picture of the scenery surface according to the triangular method and sensing structure parameter by the picture , furthermore it can calculate out three-dimensional coordinate value. The value of each point in the depth picture shows the absolute distance of every point to the camera on the scenery surface, so it can directly get the form and position message of the target from that.
This kind of method has two main advantages. First: It has formed a light mode, which has already been known in the work space, the light mode’s changing indicates that the object is changing. It can also reduce the calculating complication、improve three-dimension range measurement precision. Furthermore, the sensor based on this method has the advantages of small, light, low price, easy installed used and maintained, this makes it become the most frequently used technology at the three-dimension apperception in industry. But the method has the problem of sheltering, precision and speed contradict each other and they are difficult to raise meanwhile.
structured-light method can be divided into single method , mongline method and pattern method according to the kind of project light pattern. Namely they are point, one-dimensional line and two-dimensional pattern. Thereupon the time of the scanning can be shortened, but the corresponding degree of difficulty of confirming corresponding relation will increase.
At present people commonly analyze and measure the 3D shape of precious stones by hand in the course of its machining process. The process of measurement is complex and difficult. What’s more, it’s hard to grasp the characteristics of the main body in the process of designing the product’s figure based on the application requirement. The material will be wasted easily and the efficiency is low. In order to solve these problems, this text deploys research to find a 3D measurement system of precious stones. It also designs a measurement system of precious stones’3D shape based on structured light. The
content of this text is as follows:
Chapter 1:Prodrome. It introduces the investigation status of the machine vision in the country and abroad and its application instance in some fields. It also illustrates the purpose and meaning of this text.
Chapter 2:The structured light 3D measurement theory. This chapter mainly discusses the imaging model of the vidicon, the demarcating of the vidicon and the 3D measurement model based on the linear structured light. At first we introduce the imaging model of the object. Then we expound the method of the linear model demarcating which can ascertain the correspondence between the reference frame of the object and the reference frame of the picture. Then we introduce the theory of the structured-light inspection and measurement. At last we found the 3D measurement model based on the linear structured light and give the formulas of founding the model.
Chapter 3:pretreatment of structured light image. Original pictures obtained from the vidicon are 2D ones. They have no depth information. In order to obtain the 3D depth information from these pictures, we distill the useful 2D information from the original picture to get the 3D depth. So we deal the picture with the two-value treatment at first. This paper compare with the two ways getting threshold and use the p-tile method to distill the threshold according to the requirement of the system at last. The worked picture may have the yawp or the unclear brim, so we deal the picture with filtrating the wave by some methods. After comparing with these methods, we choose the keeping brim filter. Then we distill the skeleton of the figure being irradiated by scanning every pels one by one.
Chapter 4:system design of 3D measurement. This chapter designs a system to measure the 3D coordinate value of precious stones based on the 3D measurement math model which is established on Chapter Two and is based on the structured light. This system includes software system and hardware system. The software design of the system which is based on the software requirement analysis and the blocking structure requirement includes 5 big blocks of the system. The 5 big blocks are: the block of obtaining, the original pictures, the block of electrical engine controlment, the block of image manipulation, the block of picking up the 3D coordinate and the block of 3D repeated formation. It empolders a simple and convenient software interface based on the function of the blocks. It also provides the main program sound code. The hardware design of the system includes a box which can satisfy the measurement precision with good real time and simple operation. This
本文介绍了结构光三维视觉原理, 并在广义坐标系中采用透视投影变换理论建立了简单合理的基于线结构光的三维测量数学模型,并根据宝石的物理特性设计了线结构光三维视觉测量系统。最后用一个高精度的已知其半径与高度的轴承圆柱滚动体作为实验物进行三维测量,测量的数据走向与条纹的形状是符合的,因此说整个系统的设计理念是正确的。
structured-light method is to transform the light beam that the laser instrument is sent out into different geometry shape light through the optical system, including point, mongline, multithreading,single round , concentric more round,gridding cross etc. casts to scenery, forms the pattern and ingested by the camera on the scenery (this method is called initiative sensing method), and then calculates , gets the depth picture of the scenery surface according to the triangular method and sensing structure parameter by the picture , furthermore it can calculate out three-dimensional coordinate value. The value of each point in the depth picture shows the absolute distance of every point to the camera on the scenery surface, so it can directly get the form and position message of the target from that.
This kind of method has two main advantages. First: It has formed a light mode, which has already been known in the work space, the light mode’s changing indicates that the object is changing. It can also reduce the calculating complication、improve three-dimension range measurement precision. Furthermore, the sensor based on this method has the advantages of small, light, low price, easy installed used and maintained, this makes it become the most frequently used technology at the three-dimension apperception in industry. But the method has the problem of sheltering, precision and speed contradict each other and they are difficult to raise meanwhile.
structured-light method can be divided into single method , mongline method and pattern method according to the kind of project light pattern. Namely they are point, one-dimensional line and two-dimensional pattern. Thereupon the time of the scanning can be shortened, but the corresponding degree of difficulty of confirming corresponding relation will increase.
At present people commonly analyze and measure the 3D shape of precious stones by hand in the course of its machining process. The process of measurement is complex and difficult. What’s more, it’s hard to grasp the characteristics of the main body in the process of designing the product’s figure based on the application requirement. The material will be wasted easily and the efficiency is low. In order to solve these problems, this text deploys research to find a 3D measurement system of precious stones. It also designs a measurement system of precious stones’3D shape based on structured light. The
content of this text is as follows:
Chapter 1:Prodrome. It introduces the investigation status of the machine vision in the country and abroad and its application instance in some fields. It also illustrates the purpose and meaning of this text.
Chapter 2:The structured light 3D measurement theory. This chapter mainly discusses the imaging model of the vidicon, the demarcating of the vidicon and the 3D measurement model based on the linear structured light. At first we introduce the imaging model of the object. Then we expound the method of the linear model demarcating which can ascertain the correspondence between the reference frame of the object and the reference frame of the picture. Then we introduce the theory of the structured-light inspection and measurement. At last we found the 3D measurement model based on the linear structured light and give the formulas of founding the model.
Chapter 3:pretreatment of structured light image. Original pictures obtained from the vidicon are 2D ones. They have no depth information. In order to obtain the 3D depth information from these pictures, we distill the useful 2D information from the original picture to get the 3D depth. So we deal the picture with the two-value treatment at first. This paper compare with the two ways getting threshold and use the p-tile method to distill the threshold according to the requirement of the system at last. The worked picture may have the yawp or the unclear brim, so we deal the picture with filtrating the wave by some methods. After comparing with these methods, we choose the keeping brim filter. Then we distill the skeleton of the figure being irradiated by scanning every pels one by one.
Chapter 4:system design of 3D measurement. This chapter designs a system to measure the 3D coordinate value of precious stones based on the 3D measurement math model which is established on Chapter Two and is based on the structured light. This system includes software system and hardware system. The software design of the system which is based on the software requirement analysis and the blocking structure requirement includes 5 big blocks of the system. The 5 big blocks are: the block of obtaining, the original pictures, the block of electrical engine controlment, the block of image manipulation, the block of picking up the 3D coordinate and the block of 3D repeated formation. It empolders a simple and convenient software interface based on the function of the blocks. It also provides the main program sound code. The hardware design of the system includes a box which can satisfy the measurement precision with good real time and simple operation. This
引文
【1】段峰,王耀南,雷晓峰.机器视觉技术及其应用综述,自动化博览,2002, 19(3).pp59～61.
【2】刘曙光,刘明远.机器视觉及其应用.机械制造.2000. 38 (7) . pp20～22.
【3】荆仁杰,叶秀清,徐胜荣,陈存椿.计算机图像处理.浙江大学出版社.1990.6.
【4】史桂蓉, 邢渊,张永清.反向工程数据获取和几何建模方法.机械科学与技术. 2000年S1 期. pp 161～163.
【5】吴剑波, 李明鸣, 赵宏, 谭玉山. 立体视觉三维测量系统中的数据获取技术.仪器仪表学报. 2001 年S2 期. pp 216～218
【6】阳道善, 陈吉红, 周会成.线结构激光-机器视觉三角测量光路设计. 光学技术2001 年02 期. pp 164～168.
【7】龙玺, 钟约先, 李仁举.由志福. 结构光三维扫描测量的三维拼接技术. 清华大学学报(自然科学版) 2002 年04 期. pp 476～480.
【8】郭东明,乌秀春,王晓明, 周伟峰. 反求工程中基于图像灰度信息的三维曲面重构. 机械工程学报2003 年01 期. pp 47～50.
【9】朱洲,李德华, 关景火. 结构光测距系统仿真研究. 计算机仿真2004 年08 期. pp 132～134.
【10】李维诗.基于医学断层轮廓数据的反求CAD 建模理论与方法研究[博士学位论文]1 杭州: 浙江大学, 2001.
【11】朱洲, 李德华,关景火. 结构光测距系统仿真研究. 计算机仿真2004年08期:pp132～135.
【12】胡斌,李德华,金刚等. 基于方向模板的结构光条纹中心检测方法. 计算机工程与应用.2002 , 38 (11) : pp 59～61.
【13】马颂德,张正友.计算机视觉.北京.科学出版社.1998年.
【14】刘志刚, 方勇等. 线结构光三维视觉传感器的一种B 样条神经网络模型. 机械科学与技术, 1999, 18 (1):pp 61～78.
【15】焦李成L 神经网络系统理论[M ]L 西安: 西安电子科技大学出版社, 1990L 12L.
【16】毛剑飞, 邹细勇, 诸静. 改进的平面模板两步法标定摄像机. 中国图象图形学报2004 年07 期. pp846～850.
【17】程宏煌,戴卫恒,姚甦甦. 图像分割方法综述. 电信快报2000 年10 期. pp 39～42.
【18】Castleman KR. D igital Image P rocessing. 北京: 清华大学出版社, 1998: pp 207～250.
【19】谢毅, 陈学佺.一种实用的二值图像细化算法. 计算机应用与软件2001 年09 期. pp 21～28
【20】赵毅. 数字滤波的滑动平均法和低通滤波法. 仪表技术2001 年05 期pp 40～43 .
【21】高浩军, 杜宇人. 中值滤波在图像处理中的应用. 电子工程师2004 年08 期. pp 35～37.
【22】舒志龙, 阮秋琦. 一种二维均值滤波快速算法及其应用.2001(4). pp22～26.
【23】赵乘麟.一种新的保边缘滤波算法. 电子与信息学报2003年11期pp 1581～1584.
【24】吕俊白. 一种有效的二值图像细化算法. 计算机工程2003 年18 期pp 147～148.
【25】张远鹏,董海, 海周文灵编著。计算机图像处理技术基础。北京大学出版社,1996-09.
【26】劳俊, 伍世虔,杨叔子. 模块化与现代制造技术. 制造技术与机床1994年09期. pp 45～48.
【27】伍世虔, 孙炜, 苈俊. 计算机辅助模块化设计. 机械设计1994年04期. pp 23～27.
【28】孙峰, 张本恕. 头核磁共振鉴别诊断帕金森综合征. 天津医药2004年07期. pp 428～430.
【29】熊晓红, 王从军, 黄树槐.三维激光扫描反求系统的研究.机械与电子2004 年07 期. pp 18～20.
【30】龚志远. 付伟. 利用三坐标测量“点云”的曲面反求设计. 华东交通大学学报2004 年05 期. pp 113～116.
【31】王隆太. 机械CAD/ CAM 技术[M] . 北京:机械工业出版社,2001
【32】刘志刚,陈康宁, 刘布昆,林志航.逆向工程中基于线结构光视觉传感器的光学坐标测量系统研究. 制造业自动化1999 年05 期pp 36～39.
【33】贺忠海, 王宝光.线结构光传感器的模型及成像公式. 光学精密工程2001 年03期. pp 269～272.
【34】尤政,胡庆英,等. 结构光传感器的成像分析及其数学模型[J ] . 宇航计测技术.1997,17(4). pp 5～8.
【35】李瑞君, 范光照, 吴彰良. 线结构光传感器的精度分析及优化设计. 合肥工业大学学报(自然科学版) 2004 年10 期. pp 1115～1118.
【36】郭绍仲。透视变换矩阵。天津大学学报.1990(1). pp 67～68.
【37】张吉华等. 激光三角法轮廓测量[J ]. 天津大学学报, 1995, 28 (2) pp 265～269.
【38】赵梅芳, 沈邦兴, 吴晓明. 多目立体视觉测量系统中的快速去伪匹配法. 中国计量学院学报2003 年02 期. pp 89～93.
【39】张超, 沈振康, 张淑琴. 基于立体视觉的测量技术. 系统工程与电子技术. 24:9. pp 126～129.
【40】桑新柱, 董明利,邓文怡, 吕乃光. 四站位三维视觉测量系统. 光电子.激光2002年11 期. pp 1152～1154.
【41】陈棣湘, 罗飞路, 朱霞辉. 自由曲面视觉测量中一种有效的粗差剔除方法. 计算机应用研究1997 年05 期. pp 40～44.
【42】牛小兵, 林玉池,赵美蓉. 光栅投影三维测量的原理及关键技术分析. 仪器仪表学报2001 年S2 期. pp 203～206.
【43】A. Fusiello. Uncalibrated Euclidean reconstruction: a review. Image and
【2】刘曙光,刘明远.机器视觉及其应用.机械制造.2000. 38 (7) . pp20～22.
【3】荆仁杰,叶秀清,徐胜荣,陈存椿.计算机图像处理.浙江大学出版社.1990.6.
【4】史桂蓉, 邢渊,张永清.反向工程数据获取和几何建模方法.机械科学与技术. 2000年S1 期. pp 161～163.
【5】吴剑波, 李明鸣, 赵宏, 谭玉山. 立体视觉三维测量系统中的数据获取技术.仪器仪表学报. 2001 年S2 期. pp 216～218
【6】阳道善, 陈吉红, 周会成.线结构激光-机器视觉三角测量光路设计. 光学技术2001 年02 期. pp 164～168.
【7】龙玺, 钟约先, 李仁举.由志福. 结构光三维扫描测量的三维拼接技术. 清华大学学报(自然科学版) 2002 年04 期. pp 476～480.
【8】郭东明,乌秀春,王晓明, 周伟峰. 反求工程中基于图像灰度信息的三维曲面重构. 机械工程学报2003 年01 期. pp 47～50.
【9】朱洲,李德华, 关景火. 结构光测距系统仿真研究. 计算机仿真2004 年08 期. pp 132～134.
【10】李维诗.基于医学断层轮廓数据的反求CAD 建模理论与方法研究[博士学位论文]1 杭州: 浙江大学, 2001.
【11】朱洲, 李德华,关景火. 结构光测距系统仿真研究. 计算机仿真2004年08期:pp132～135.
【12】胡斌,李德华,金刚等. 基于方向模板的结构光条纹中心检测方法. 计算机工程与应用.2002 , 38 (11) : pp 59～61.
【13】马颂德,张正友.计算机视觉.北京.科学出版社.1998年.
【14】刘志刚, 方勇等. 线结构光三维视觉传感器的一种B 样条神经网络模型. 机械科学与技术, 1999, 18 (1):pp 61～78.
【15】焦李成L 神经网络系统理论[M ]L 西安: 西安电子科技大学出版社, 1990L 12L.
【16】毛剑飞, 邹细勇, 诸静. 改进的平面模板两步法标定摄像机. 中国图象图形学报2004 年07 期. pp846～850.
【17】程宏煌,戴卫恒,姚甦甦. 图像分割方法综述. 电信快报2000 年10 期. pp 39～42.
【18】Castleman KR. D igital Image P rocessing. 北京: 清华大学出版社, 1998: pp 207～250.
【19】谢毅, 陈学佺.一种实用的二值图像细化算法. 计算机应用与软件2001 年09 期. pp 21～28
【20】赵毅. 数字滤波的滑动平均法和低通滤波法. 仪表技术2001 年05 期pp 40～43 .
【21】高浩军, 杜宇人. 中值滤波在图像处理中的应用. 电子工程师2004 年08 期. pp 35～37.
【22】舒志龙, 阮秋琦. 一种二维均值滤波快速算法及其应用.2001(4). pp22～26.
【23】赵乘麟.一种新的保边缘滤波算法. 电子与信息学报2003年11期pp 1581～1584.
【24】吕俊白. 一种有效的二值图像细化算法. 计算机工程2003 年18 期pp 147～148.
【25】张远鹏,董海, 海周文灵编著。计算机图像处理技术基础。北京大学出版社,1996-09.
【26】劳俊, 伍世虔,杨叔子. 模块化与现代制造技术. 制造技术与机床1994年09期. pp 45～48.
【27】伍世虔, 孙炜, 苈俊. 计算机辅助模块化设计. 机械设计1994年04期. pp 23～27.
【28】孙峰, 张本恕. 头核磁共振鉴别诊断帕金森综合征. 天津医药2004年07期. pp 428～430.
【29】熊晓红, 王从军, 黄树槐.三维激光扫描反求系统的研究.机械与电子2004 年07 期. pp 18～20.
【30】龚志远. 付伟. 利用三坐标测量“点云”的曲面反求设计. 华东交通大学学报2004 年05 期. pp 113～116.
【31】王隆太. 机械CAD/ CAM 技术[M] . 北京:机械工业出版社,2001
【32】刘志刚,陈康宁, 刘布昆,林志航.逆向工程中基于线结构光视觉传感器的光学坐标测量系统研究. 制造业自动化1999 年05 期pp 36～39.
【33】贺忠海, 王宝光.线结构光传感器的模型及成像公式. 光学精密工程2001 年03期. pp 269～272.
【34】尤政,胡庆英,等. 结构光传感器的成像分析及其数学模型[J ] . 宇航计测技术.1997,17(4). pp 5～8.
【35】李瑞君, 范光照, 吴彰良. 线结构光传感器的精度分析及优化设计. 合肥工业大学学报(自然科学版) 2004 年10 期. pp 1115～1118.
【36】郭绍仲。透视变换矩阵。天津大学学报.1990(1). pp 67～68.
【37】张吉华等. 激光三角法轮廓测量[J ]. 天津大学学报, 1995, 28 (2) pp 265～269.
【38】赵梅芳, 沈邦兴, 吴晓明. 多目立体视觉测量系统中的快速去伪匹配法. 中国计量学院学报2003 年02 期. pp 89～93.
【39】张超, 沈振康, 张淑琴. 基于立体视觉的测量技术. 系统工程与电子技术. 24:9. pp 126～129.
【40】桑新柱, 董明利,邓文怡, 吕乃光. 四站位三维视觉测量系统. 光电子.激光2002年11 期. pp 1152～1154.
【41】陈棣湘, 罗飞路, 朱霞辉. 自由曲面视觉测量中一种有效的粗差剔除方法. 计算机应用研究1997 年05 期. pp 40～44.
【42】牛小兵, 林玉池,赵美蓉. 光栅投影三维测量的原理及关键技术分析. 仪器仪表学报2001 年S2 期. pp 203～206.
【43】A. Fusiello. Uncalibrated Euclidean reconstruction: a review. Image and
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |