单目视觉坐标测量系统建模的研究
|
摘要
随着世界科技的迅速发展,特别是航空、航天、军工等传统大尺寸工业都已对大尺寸测量提出了新的要求,传统的大尺寸测量方法已无法满足现代化生产对空间三维尺寸的测量要求,需要找到一种可以在生产、装配过程中进行校准的大尺寸三维测量技术,而视觉坐标测量技术作为一种新兴三维空间坐标测量技术,从一个全新的坐标测量思路上适应了现代化生产对测量技术在测量精度、测量范围等方面新的更高要求。
本文以建立单目视觉坐标测量系统的模型为目的,研究了测头成像式单目视觉三维坐标测量技术,采用基于几何和基于图像的混合建模方法对单目视觉坐标测量系统进行建模。论文简要分析了测量系统的硬件组成及工作原理,详细地给出了模型的建立过程和求解过程并通过基于空间坐标约束的位姿优化算法确定被测点(测尖)的空间坐标;针对摄像机的非线性模型,摄像机在成像过程中会出现畸变问题,为了使对应像点能够更加接近摄像机线性模型中的理想像点,本文采用了利用交比不变性的原理来求解畸变参数,进而实现对图像像点的校正;针对测头上光学特征点的发光特性,会使对应像点中心位置很难确定,在此采用了高斯双三次插值算法对像点进行了质心定位研究;进行了图像像点校正实验、质心定位实验和系统测量模型的验证实验,并对产生系统模型精度及稳定性的误差因素进行了分析。
With the rapid development of science and technology,the appearance design of large and giant product in Aviation, aerospace, military and civilian areas are more and more complex. And its processing and assembling have become increasingly demanding high precision and in most cases real-time requirements of the scene. So the traditional Measurement of the content and methods have been unable to meet the three-dimensional space measurement requirements of the modern production. There is an urgent need to find a ladge-size three-dimensional overall field test technology and system which can carry out three-dimensional high-precision structure measurement to large pieces of weapons and equipment in the scene of production, assemble and use, such as missiles, satellites, aircraft, tanks and so on. While Vision coordinate measurement technology is a new three-dimensional space coordinate technology based on computer vision and coordinate measuring technology research, which has adopt to higher demand of the measuring technology of the modern product in aspects of measuring accuracy, measuring range from a new thinking of coordinate measurement.
Basing on vision coordinate measurement technology, this thesis poses monocular vision coordinate measurement system on the base of measurement probe imaging.This method has changed the previous condition that vision coordinate measurement instrument is limited by acquiring and processing the measured appearance directly, and also changed the condition of limit application because of complex shape of measured object. So it is fit for online measurement of industrial site. For the whole vision measurement system, the primary task is to set up the system modle, which is also the key influencing factor to measurement system precision. Here,we focus on sovling the problem of system modleing of monocular vision coordinate measurement.
The system is maily composed of high Resolution black and white CCD camera, a lens with infrared high-pass filter, a portable industrial computer, image acquisition and processing system, a RS-485 serial communication interface and so on. Through the 5 analysis of optical characteristic points’imaging coordinate on optical measurement probe, making use of knowledge of geometric constraint among every characteristic point we know, and linear calibration of camera nonlinear modle, we can fix the characteristic points’relative coordinates to camera coordinate system, and fix the measured points’space coordinates by making use of the position relation between probe tip(contact) and the characteristic points, then realize all functions of coordinate measurement instrument.
This thesis set up the system model through perspective projection camera model corrected by making use of two-dimensional distortion image, image coordinates with optical characteristics and measurement principle of optical probe’s geometric location constraint knowledge. Also it gives out the specific solution process, and fix the space coordinates of measured points(probe tip) based on pose optimization algorithm of space coordinate constraint. In order to make the model practicle, we need to increase the system model’s precision and stability. So we pose all kinds of factors which could influence precision and stability of the system model, and give out the specific solution. Then we introduce imaging model of camera, realize that the camera is a nonlinear model with distortion during actual imaging process. In order to solve the problem of that, and make the image point more close to ideal image point of camera linear model, also decrease the error of system, increase demarcate precision, through the analysis of the fomer nonlinear distortion correction method, only considering radial distortion of lens, we pose nonlinear distortion correction method based on cross-ratio invariance, which is a demarcate method independent to other internal and external prarmeters, with simple calculation, fast and accurate. And because of instability factor existing during the process of probe tip pre-calibration and replacing, we must calibrate the probe tip coordinates accurately before erery measurement in order to make its influence to the system model decrease to least. Thus, we adopt the constraints of camera’s coordinates no changing during the swing process for the probe tip fixed in spherical seat to correct the probe tip’s coordinates. Then, for the characteristic of optical characteristics point in probe tip, we do the research of centroid position to the image point of characteristics point. Considering the Gauss fitting’s better anti-noise ability and higher precision compared to centroid method, we adpot Gauss Bi-cubic interpolation algorithm to solve the centroid position problem in monocular vision coordinate measurement system. But it has only made use of imaging edge points of characteristics points, ignoring the other points’function on the whole characteristics points imaging surface. So, basing on traditional Gauss fitting algorithm, we adpot interpolation algorithm to increase the amount of effective pixels points around the characteristics points, and improve the location precision of characteristics points’center through least square method to eliminate error factors. Thereinto, bicubic interpolation has increased 16 effective pixels around characteristics points, more than bilinear interpolation, and it could improve the precision. Finally, it makes verified test of system parameter correction, centroid location and system measurement model by making use of self-made plane target and Matlab program, and also analyses the error factors causing precision and stability of system model. Under the actual experiment condition, the experiment indicates that the system model set up by this thesis only could be used in site measurement with not high precision demand.
本文以建立单目视觉坐标测量系统的模型为目的,研究了测头成像式单目视觉三维坐标测量技术,采用基于几何和基于图像的混合建模方法对单目视觉坐标测量系统进行建模。论文简要分析了测量系统的硬件组成及工作原理,详细地给出了模型的建立过程和求解过程并通过基于空间坐标约束的位姿优化算法确定被测点(测尖)的空间坐标;针对摄像机的非线性模型,摄像机在成像过程中会出现畸变问题,为了使对应像点能够更加接近摄像机线性模型中的理想像点,本文采用了利用交比不变性的原理来求解畸变参数,进而实现对图像像点的校正;针对测头上光学特征点的发光特性,会使对应像点中心位置很难确定,在此采用了高斯双三次插值算法对像点进行了质心定位研究;进行了图像像点校正实验、质心定位实验和系统测量模型的验证实验,并对产生系统模型精度及稳定性的误差因素进行了分析。
With the rapid development of science and technology,the appearance design of large and giant product in Aviation, aerospace, military and civilian areas are more and more complex. And its processing and assembling have become increasingly demanding high precision and in most cases real-time requirements of the scene. So the traditional Measurement of the content and methods have been unable to meet the three-dimensional space measurement requirements of the modern production. There is an urgent need to find a ladge-size three-dimensional overall field test technology and system which can carry out three-dimensional high-precision structure measurement to large pieces of weapons and equipment in the scene of production, assemble and use, such as missiles, satellites, aircraft, tanks and so on. While Vision coordinate measurement technology is a new three-dimensional space coordinate technology based on computer vision and coordinate measuring technology research, which has adopt to higher demand of the measuring technology of the modern product in aspects of measuring accuracy, measuring range from a new thinking of coordinate measurement.
Basing on vision coordinate measurement technology, this thesis poses monocular vision coordinate measurement system on the base of measurement probe imaging.This method has changed the previous condition that vision coordinate measurement instrument is limited by acquiring and processing the measured appearance directly, and also changed the condition of limit application because of complex shape of measured object. So it is fit for online measurement of industrial site. For the whole vision measurement system, the primary task is to set up the system modle, which is also the key influencing factor to measurement system precision. Here,we focus on sovling the problem of system modleing of monocular vision coordinate measurement.
The system is maily composed of high Resolution black and white CCD camera, a lens with infrared high-pass filter, a portable industrial computer, image acquisition and processing system, a RS-485 serial communication interface and so on. Through the 5 analysis of optical characteristic points’imaging coordinate on optical measurement probe, making use of knowledge of geometric constraint among every characteristic point we know, and linear calibration of camera nonlinear modle, we can fix the characteristic points’relative coordinates to camera coordinate system, and fix the measured points’space coordinates by making use of the position relation between probe tip(contact) and the characteristic points, then realize all functions of coordinate measurement instrument.
This thesis set up the system model through perspective projection camera model corrected by making use of two-dimensional distortion image, image coordinates with optical characteristics and measurement principle of optical probe’s geometric location constraint knowledge. Also it gives out the specific solution process, and fix the space coordinates of measured points(probe tip) based on pose optimization algorithm of space coordinate constraint. In order to make the model practicle, we need to increase the system model’s precision and stability. So we pose all kinds of factors which could influence precision and stability of the system model, and give out the specific solution. Then we introduce imaging model of camera, realize that the camera is a nonlinear model with distortion during actual imaging process. In order to solve the problem of that, and make the image point more close to ideal image point of camera linear model, also decrease the error of system, increase demarcate precision, through the analysis of the fomer nonlinear distortion correction method, only considering radial distortion of lens, we pose nonlinear distortion correction method based on cross-ratio invariance, which is a demarcate method independent to other internal and external prarmeters, with simple calculation, fast and accurate. And because of instability factor existing during the process of probe tip pre-calibration and replacing, we must calibrate the probe tip coordinates accurately before erery measurement in order to make its influence to the system model decrease to least. Thus, we adopt the constraints of camera’s coordinates no changing during the swing process for the probe tip fixed in spherical seat to correct the probe tip’s coordinates. Then, for the characteristic of optical characteristics point in probe tip, we do the research of centroid position to the image point of characteristics point. Considering the Gauss fitting’s better anti-noise ability and higher precision compared to centroid method, we adpot Gauss Bi-cubic interpolation algorithm to solve the centroid position problem in monocular vision coordinate measurement system. But it has only made use of imaging edge points of characteristics points, ignoring the other points’function on the whole characteristics points imaging surface. So, basing on traditional Gauss fitting algorithm, we adpot interpolation algorithm to increase the amount of effective pixels points around the characteristics points, and improve the location precision of characteristics points’center through least square method to eliminate error factors. Thereinto, bicubic interpolation has increased 16 effective pixels around characteristics points, more than bilinear interpolation, and it could improve the precision. Finally, it makes verified test of system parameter correction, centroid location and system measurement model by making use of self-made plane target and Matlab program, and also analyses the error factors causing precision and stability of system model. Under the actual experiment condition, the experiment indicates that the system model set up by this thesis only could be used in site measurement with not high precision demand.
引文
[1]马骊群,王立鼎等.工业大尺寸测量仪器的溯源现状及发展趋势[J].计测技术,2006(06)
[2]周秀云,杨春生,张涛.高精度无导轨大尺寸测量系统[J].实用测试技术,2002,6,14-16
[3]周富强,张广军,江洁等.现场双经纬仪三维坐标测量系统[J].机械工程学报,2004
[4]祝世平,强锡富.工件特征点三维坐标视觉测量方法综述[J].光学精密工程,2000,8(2),1922197
[5]李莉,劳奇成,李平等.坐标测量新技术及发展趋势[J].西安工业学院学报,2002,18(4),36-38
[6]张之江,车仁生等.仿生计量学的新领域——视觉坐标测量机[M].宇航计测技术,1998
[7] D Liebowitz and A Zisserman.Metric rectification for perspective images of planes[J].Proc.of CVPR,pp:482-488,Jun.2003
[8]黄凤山.光笔式单摄像机三维坐标视觉测量系统关键技术的研究[D].天津大学博士学位论文,2005
[9]马国松.基于单目视觉的位姿测量方法及仿真系统研究[D].哈尔滨工业大学硕士学位论文,2005
[10]赵锐.基于单目视觉的物体位姿测量方法研究[D].合肥工业大学硕士学位论文,2005
[11]黄桂平,李广云,王保丰等,单目视觉测量技术研究[J],计量学报,2004,25(4),314-317
[12] G.H.Wang,Z.Y.Hu and F.C.Wu,Single view based measurement on space planes[J].Journal of Computer Science and Technology,Vol.19(3),2004
[13] Q.Long,Z.D.Lan.Linear N-Point Camera Pose Determination[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1999,21(8),774-780
[14] A Criminisi,Accurate visual metrology from single and multiple uncalibrated images-distinguished dissertations[J].Springer Verlag Press,2001
[15] A Criminisi,I Reid and A Zisserman,A plane measuring device[J].Image and Vision Computing,vol.17(8),625-634,1999
[16] G.H.Wang,Y.H.Wu and Z.Y.Hu,A novel approach for single view Based planemetrology [J].In:Proc.of the 16th International Conference on Patern Recognition,vol.2,556-559, 2002
[17]杨克俭,刘舒燕,陈定方.虚拟现实中的建模方法[J].武汉工业大学学报,2001(6),vol.6
[18] F A van den Heuvel,3D reconstruction from a single image using geometric constraints[J].Journal of Photogrammetry&Remote Sensing,pp:354-368,1998
[19]吴炜煌,谢成林.基于照片的混合建模系统几何模型求解原理研究[J].中国图象图形学报,2000
[20] Kanade T,et al.A stereo matching algorithm with an adaptive window:Theory and Experiment[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2004,16(9),920-932
[21]王春莉,梁立波,王宝玉.计算机三维重建技术发展与应用[J].沈阳大学学报,2003
[22]李辉.一种基于零件加工特征的几何建模方法[J].石家庄铁道学院学报,2006(2),77-81
[23]李成军,马志刚,汪国平等.交互的极线几何建模方法[J].计算机辅助设计与图形学学报,2002.9
[24] Rocchini,Cignoni P,Montani M.A Low Cost 3D Scanner Based on Structured Light[J].EG 2001 Proc.2001,20(3),299-308
[25] Juergen Peipe,Peter Andrae.Optical 3D Coordinate Measurement using a Range Sensor and Photogrammetry.Proceeding of Videometrics VII,Santa Clara,CA,USA,2003[J]. Proceedings SPIE.2003,5013:110-116
[26] Depiero F W,Trivedi M M.3D Computer Vision using Structured Light:Design,Calibration and Implementation Issues[J].Advances in Computers,1996:43-49
[27]段发阶,张键新,叶声华.CCD摄像机参数标定新技术[J].计量学报,1997,18(4):294-299
[28] Martin A.Fishler,Robert C.Bolles.Random Sample Consensus:A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography[J].Communications of the ACM,1981,(6):381-395
[29]杨必武,郭晓松.摄像机镜头非线性畸变校正方法综述[J].中国图象图形学报,2005,10(3)
[30]陈辉,吕乃光.基于交比不变性的相机镜头畸变标定研究[J].北京机械工业学院学报,2006,21(1)
[31]贺俊吉,张广军,杨宪铭.基于交比不变性的镜头畸变参数标定方法[J].仪器仪表学报,2004,25(5)
[32]王永强,费建锋,毕德学等.基于交比不变性的投影仪镜头畸变参数标定方法[J].测试技术学报,2008,22(4)
[33]张之江,车仁生等.测头成像视觉坐标测量系统组成、建模及求解[J].哈尔滨工业大学学报,1999
[34]张之江,车仁生,黄庆成等.单摄像机测头成像视觉坐标测量系统建模[J].哈尔滨工业大学学报,1999
[35]刘长英,叶东,车仁生.视觉测量中光学测头姿态估计方法研究[J].哈尔滨工业大学学报,2006
[36]张之江,车仁生.视觉坐标测量中直接光学法透镜光学中心的标定[M].光学精密工程,1999
[37]刘长英.大尺寸单目视觉测量关键技术研究[D].哈尔滨工业大学博士论文,2006
[38]彭凯,张雪飞等.点阵测头成像视觉坐标测量系统的研究[M].工艺与检测,2007
[39]张之江.单摄象机测头成像视觉坐标测量技术的研究[D].哈尔滨工业大学博士论文,1998
[40]黄风山,刘书桂,彭凯.跟踪测距式三坐标视觉测量系统[M].光电工程,2006
[41]彭凯,刘书桂,黄风山.光笔式便携三维坐标视觉测量系统的建模与分析[M].计量技术,2005
[42]苏成,李冬果等.判定p3p问题正解数目的充要条件[J].计算机学报,1998,21(12),1084- 1095
[43]张之江,张智强,陈迎娟,靶标成像双目视觉坐标测量建模[J].计量学报,1999
[44]董钧祥,杨德宏.测量坐标转换模型及其应用[J].昆明理工大学学报,2006
[45]王亚平,王鹏.三维测量软件系统中坐标系的建立[J].北京航空航天大学学报,2004
[46]范强国.大尺度视觉跟踪坐标测量系统设计[D].哈尔滨工业大学硕士学位论文,2007
[47]刘志刚,方勇,陈康宁等.线结构光三维视觉曲面测量的自适应采样与建模方法[J].西安交通大学学报,1999,33(10)
[48]陈祖明.矩阵论引论[M].北京航空航天大学出版社,1998
[49]解凯,郭恒业,张田文.基于图象建模技术的综述[J].系统仿真学报,2O01
[50]张之江等.测头成像视觉坐标测量系统中特征点成像中心获取[M].光学精密工程,1998
[51]武晓杰.基于特征点距离约束的机器人视觉定位方法[D].吉林大学硕士学位论文,2008
[52]靖娜.采用光学测棒的三目视觉坐标测量系统的理论研究[D].哈尔滨工业大学硕士学位论文,2006
[53]李敏,盛毅.高斯拟合算法在光谱建模中的应用研究[M].光谱学与光谱分析,2008,28(10)
[54]罗立彦,杨勋年.基于细分的图像插值算法[J].计算机辅助设计与图形学报,2006,18(9)
[2]周秀云,杨春生,张涛.高精度无导轨大尺寸测量系统[J].实用测试技术,2002,6,14-16
[3]周富强,张广军,江洁等.现场双经纬仪三维坐标测量系统[J].机械工程学报,2004
[4]祝世平,强锡富.工件特征点三维坐标视觉测量方法综述[J].光学精密工程,2000,8(2),1922197
[5]李莉,劳奇成,李平等.坐标测量新技术及发展趋势[J].西安工业学院学报,2002,18(4),36-38
[6]张之江,车仁生等.仿生计量学的新领域——视觉坐标测量机[M].宇航计测技术,1998
[7] D Liebowitz and A Zisserman.Metric rectification for perspective images of planes[J].Proc.of CVPR,pp:482-488,Jun.2003
[8]黄凤山.光笔式单摄像机三维坐标视觉测量系统关键技术的研究[D].天津大学博士学位论文,2005
[9]马国松.基于单目视觉的位姿测量方法及仿真系统研究[D].哈尔滨工业大学硕士学位论文,2005
[10]赵锐.基于单目视觉的物体位姿测量方法研究[D].合肥工业大学硕士学位论文,2005
[11]黄桂平,李广云,王保丰等,单目视觉测量技术研究[J],计量学报,2004,25(4),314-317
[12] G.H.Wang,Z.Y.Hu and F.C.Wu,Single view based measurement on space planes[J].Journal of Computer Science and Technology,Vol.19(3),2004
[13] Q.Long,Z.D.Lan.Linear N-Point Camera Pose Determination[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1999,21(8),774-780
[14] A Criminisi,Accurate visual metrology from single and multiple uncalibrated images-distinguished dissertations[J].Springer Verlag Press,2001
[15] A Criminisi,I Reid and A Zisserman,A plane measuring device[J].Image and Vision Computing,vol.17(8),625-634,1999
[16] G.H.Wang,Y.H.Wu and Z.Y.Hu,A novel approach for single view Based planemetrology [J].In:Proc.of the 16th International Conference on Patern Recognition,vol.2,556-559, 2002
[17]杨克俭,刘舒燕,陈定方.虚拟现实中的建模方法[J].武汉工业大学学报,2001(6),vol.6
[18] F A van den Heuvel,3D reconstruction from a single image using geometric constraints[J].Journal of Photogrammetry&Remote Sensing,pp:354-368,1998
[19]吴炜煌,谢成林.基于照片的混合建模系统几何模型求解原理研究[J].中国图象图形学报,2000
[20] Kanade T,et al.A stereo matching algorithm with an adaptive window:Theory and Experiment[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2004,16(9),920-932
[21]王春莉,梁立波,王宝玉.计算机三维重建技术发展与应用[J].沈阳大学学报,2003
[22]李辉.一种基于零件加工特征的几何建模方法[J].石家庄铁道学院学报,2006(2),77-81
[23]李成军,马志刚,汪国平等.交互的极线几何建模方法[J].计算机辅助设计与图形学学报,2002.9
[24] Rocchini,Cignoni P,Montani M.A Low Cost 3D Scanner Based on Structured Light[J].EG 2001 Proc.2001,20(3),299-308
[25] Juergen Peipe,Peter Andrae.Optical 3D Coordinate Measurement using a Range Sensor and Photogrammetry.Proceeding of Videometrics VII,Santa Clara,CA,USA,2003[J]. Proceedings SPIE.2003,5013:110-116
[26] Depiero F W,Trivedi M M.3D Computer Vision using Structured Light:Design,Calibration and Implementation Issues[J].Advances in Computers,1996:43-49
[27]段发阶,张键新,叶声华.CCD摄像机参数标定新技术[J].计量学报,1997,18(4):294-299
[28] Martin A.Fishler,Robert C.Bolles.Random Sample Consensus:A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography[J].Communications of the ACM,1981,(6):381-395
[29]杨必武,郭晓松.摄像机镜头非线性畸变校正方法综述[J].中国图象图形学报,2005,10(3)
[30]陈辉,吕乃光.基于交比不变性的相机镜头畸变标定研究[J].北京机械工业学院学报,2006,21(1)
[31]贺俊吉,张广军,杨宪铭.基于交比不变性的镜头畸变参数标定方法[J].仪器仪表学报,2004,25(5)
[32]王永强,费建锋,毕德学等.基于交比不变性的投影仪镜头畸变参数标定方法[J].测试技术学报,2008,22(4)
[33]张之江,车仁生等.测头成像视觉坐标测量系统组成、建模及求解[J].哈尔滨工业大学学报,1999
[34]张之江,车仁生,黄庆成等.单摄像机测头成像视觉坐标测量系统建模[J].哈尔滨工业大学学报,1999
[35]刘长英,叶东,车仁生.视觉测量中光学测头姿态估计方法研究[J].哈尔滨工业大学学报,2006
[36]张之江,车仁生.视觉坐标测量中直接光学法透镜光学中心的标定[M].光学精密工程,1999
[37]刘长英.大尺寸单目视觉测量关键技术研究[D].哈尔滨工业大学博士论文,2006
[38]彭凯,张雪飞等.点阵测头成像视觉坐标测量系统的研究[M].工艺与检测,2007
[39]张之江.单摄象机测头成像视觉坐标测量技术的研究[D].哈尔滨工业大学博士论文,1998
[40]黄风山,刘书桂,彭凯.跟踪测距式三坐标视觉测量系统[M].光电工程,2006
[41]彭凯,刘书桂,黄风山.光笔式便携三维坐标视觉测量系统的建模与分析[M].计量技术,2005
[42]苏成,李冬果等.判定p3p问题正解数目的充要条件[J].计算机学报,1998,21(12),1084- 1095
[43]张之江,张智强,陈迎娟,靶标成像双目视觉坐标测量建模[J].计量学报,1999
[44]董钧祥,杨德宏.测量坐标转换模型及其应用[J].昆明理工大学学报,2006
[45]王亚平,王鹏.三维测量软件系统中坐标系的建立[J].北京航空航天大学学报,2004
[46]范强国.大尺度视觉跟踪坐标测量系统设计[D].哈尔滨工业大学硕士学位论文,2007
[47]刘志刚,方勇,陈康宁等.线结构光三维视觉曲面测量的自适应采样与建模方法[J].西安交通大学学报,1999,33(10)
[48]陈祖明.矩阵论引论[M].北京航空航天大学出版社,1998
[49]解凯,郭恒业,张田文.基于图象建模技术的综述[J].系统仿真学报,2O01
[50]张之江等.测头成像视觉坐标测量系统中特征点成像中心获取[M].光学精密工程,1998
[51]武晓杰.基于特征点距离约束的机器人视觉定位方法[D].吉林大学硕士学位论文,2008
[52]靖娜.采用光学测棒的三目视觉坐标测量系统的理论研究[D].哈尔滨工业大学硕士学位论文,2006
[53]李敏,盛毅.高斯拟合算法在光谱建模中的应用研究[M].光谱学与光谱分析,2008,28(10)
[54]罗立彦,杨勋年.基于细分的图像插值算法[J].计算机辅助设计与图形学报,2006,18(9)