光流计算及其在被动测距中的应用研究
|
摘要
随着科技的日新月异,测量测控技术不断的引入新的元素。图像处理技术也越来越多地应用到光学测量测距方面,使得光学测量测距技术在众多工程领域有了提高和改善。
本文重点研究了光流算法并将其应用到了被动测距中,对被测图像进行了处理,并给出光流计算的结果。
首先,本文总结了几种基本被动测距和主动测距的测量方法。并对各个方法的优缺点进行了简单的分析。设计了双目立体测距和三角法动态测速相结合的方法,并运用在本课题中。对于三角法在动态测量中的应用采用了现在比较先进的高精度PSD器件来定位光学投影中心。结合光流和差分—特征点两种方法,推导出光流与实际被测速度之间关系的一般公式。
然后,采用基本图像处理方法实现对拍摄的测量图像进行处理。分别采用了不同的阈值、边界检验、特征值提取和差分—特征点的方法实现处理,并对处理结果进行了比较,选用比较合适的算法应用到本课题中,从而得到被测物体的基本运动状态。最后,对光流算子进行了深入研究。将基于Prewiit和Hueckels的算子改进成3D算子,并对这几种算子处理的图像进行了比较,且运用该算子应用于光流计算。
With improvement of the technology, many new ideas including the image processing technology have been applied in the measurement and testing technology. This makes the optical measurement and testing technology to be improved in the various fields.
The optical flow estimation algorithm is studied in this paper and is used to the passive ranging. The result of optical flow is given by the image processing of real target.
First, the basic knowledge of passive ranging and active ranging is given and different methods are compared. The binocular passive ranging and Dynamic trigonometry ranging are modified and used in this paper. In the dynamic trigonometry ranging system the high accuracy PSD device is used to calculate the light projecting central point. Combined the results from the methods of optical flow and difference-feature, the expressions between the optical flow and the velocity of measured target is deduced.
Hence, the basic image processing method is applied to real experimental result. Comparing the results from the threshold methods, edge detecting algorithm and difference-feature, the algorithm which is suitable to our project is chosen to estimate the motion state of the target. At last, the optical flow computation algorithms are deeply studied. The Prewiit and Hueckels arithmetic operators are modified to the 3D form. Then the 3D-algorithem is applied to compute the optical flow.
本文重点研究了光流算法并将其应用到了被动测距中,对被测图像进行了处理,并给出光流计算的结果。
首先,本文总结了几种基本被动测距和主动测距的测量方法。并对各个方法的优缺点进行了简单的分析。设计了双目立体测距和三角法动态测速相结合的方法,并运用在本课题中。对于三角法在动态测量中的应用采用了现在比较先进的高精度PSD器件来定位光学投影中心。结合光流和差分—特征点两种方法,推导出光流与实际被测速度之间关系的一般公式。
然后,采用基本图像处理方法实现对拍摄的测量图像进行处理。分别采用了不同的阈值、边界检验、特征值提取和差分—特征点的方法实现处理,并对处理结果进行了比较,选用比较合适的算法应用到本课题中,从而得到被测物体的基本运动状态。最后,对光流算子进行了深入研究。将基于Prewiit和Hueckels的算子改进成3D算子,并对这几种算子处理的图像进行了比较,且运用该算子应用于光流计算。
With improvement of the technology, many new ideas including the image processing technology have been applied in the measurement and testing technology. This makes the optical measurement and testing technology to be improved in the various fields.
The optical flow estimation algorithm is studied in this paper and is used to the passive ranging. The result of optical flow is given by the image processing of real target.
First, the basic knowledge of passive ranging and active ranging is given and different methods are compared. The binocular passive ranging and Dynamic trigonometry ranging are modified and used in this paper. In the dynamic trigonometry ranging system the high accuracy PSD device is used to calculate the light projecting central point. Combined the results from the methods of optical flow and difference-feature, the expressions between the optical flow and the velocity of measured target is deduced.
Hence, the basic image processing method is applied to real experimental result. Comparing the results from the threshold methods, edge detecting algorithm and difference-feature, the algorithm which is suitable to our project is chosen to estimate the motion state of the target. At last, the optical flow computation algorithms are deeply studied. The Prewiit and Hueckels arithmetic operators are modified to the 3D form. Then the 3D-algorithem is applied to compute the optical flow.
引文
[1]Baribeau R,Rioux M.Influence of speckle on laser range finders[J].Appl Opt,1991,(3):2873-2878.
[2]赵天宇,吴易明,陈良益,朱立峰,刘宇波.结构光法和相位移法的对比研究[J].科学技术与工程[J],2007,7(17):4429-4432.
[3]赵勋杰,李成金.双目立体实时测距系统的关键技术研究[J].激光与红外,2006,36(9):874-877.
[4]金元柳.对激光近炸引信的干扰技术[J].光电技术应用,2004,19(5):7-10.
[5]赵易良.立体视觉与运动分析融合问题的研究[J].光电技术应用,2007:10-20.
[6]M.Bertozzi,A.Broggi,A.Fascioli.Stereo Vision-based vehicle detection[J].IEEE,2000,pp:39-44.
[7]吴利予,李金,吕普轶,李想.AVI文件的软件解析及自动单帧捕获[J].应用科技,2003,30(9):14-15.
[8]陶琳.基于光流技术的图像信息提取[J].华中科技大学学报,2006:30-55.
[9]J.L.Barron,D.J.Fleet,S.S.Beauchemin.Performance of Optical Flow Techniques[J].Computer Vision,1994,12(1):43-77.
[10]Milan Sonka,Vaclav Hlavac,Roger Boyle.Computer Vision[M].北京:人民邮电出版社[M],2006:466-477.
[11]陈兵旗,孙明.VC++使用图像处理专业教程[M].北京:清华大学出版社,2005:19-77.
[12]Ming Ye.Robust Visual Motion Analysis:Piecewise-Smooth Optical Flow and Motion-based Detection and Tracking[J].University of Washington,2002[0]
[13]Ronald J,Pieper,Ronald J,Pieper,G.Pelegris.Passive range estimation using dual-baseline Triangulation[J].Optical Engineering,1996,35(3):685-692.
[14]李耀辉,刘宝军.运动模糊图像的光流场分析[M].华北科技学院学报,2004,1(3):86-87.
[15]陈伏虎,罗雪松,陈洪兴,翁伟,费腾.双活动平台被动测距方法研究[J].声学与电子工程,2004,3(79):15-16.
[16]危水根,陈震,黎明.基于梯度光流法计算的一种改进[J].计算机工,2006,32(1):198-200.
[17]胡金华,吴文全.被动测距误差分析及修正[J].电子测量技术,2006,29(3):38-40.
[18]Yair Barniv.Error.Analysis of Combined Optical-Flow and Stereo Passive Ranging[J].Journal of the IEEE,1992,28(4):978-989.
[19]B.Sridhar,P.Smith,R.Suorsa,B.Hussien.Multirate and Event Driven Kalman Filters for Helicopter Passive Ranging[J].IEEE Control Systens Magazine,10(4):43-50.
[20]Yair Barniv.Passive Ranging Using Image Expansion[J].IEEE,1995,31(1):358-354.
[21]Maarten de Visser,Maarten de Visser,Maarten de Visser,Emile A.Hendriks.SPIE,2006,45(2):026402-1-14.
[22]陶琳,黄士科,张天序.基于光流的红外目标被动测距技术[J].计算机与数字工程,2005,33(11):1-4.
[23]黄士科,夏涛,张天序.基于红外图像的被动测距方法[J].红外与激光工程,2007,6(1):109-126.
[24]田旭,胡桂琴.AVI格式文件的分解处理[J],河北工业科技,2006,23(4):240-243.
[25]于晓康,柴乔林.基于P2nn的可控视频回放系统.计算机应用,2003,23:326-329.
[26]Inseok Hwang,Hamsa Balakrishnan,Kaushik Royt,and Claire Tomlin.Multiple-Target Tracking and Identity Management in Clutter,with Application to Aircraft Tracking.Proceeding of the 2004 American Control Conference,2004,1:3422-3428.
[27]Dr.Paul Frank Singer,Doreen M.Sasaki.Analysis of the Cascade of Track-Before-Detect and Track-After-Detect Tracking Algorithms[J].SPIE,1998,3373:157-165.
[28]Boris Rozovskii,Anton PetrovOptimal.Nonlinear filtering for track—before-detect in JR image sequences[J].SPIE,1999,3809:152-163.
[29]郭丰俊,杨新,W.Phliips,施鹏飞,I.Lemahieu.基于HUBER函数的序列图像位移场估计算法[J].高技术通讯,2001,34:38-42.
[30]白雪生,金国,徐光祜,史元春.基于帧间运动的视频分割[J].清华大学学报(自然科学版),2002,42(1):76-79。
[31]李熙莹,倪国强.红外图像的光流计算[J].红外与激光工程,2002,31(3):189-193.
[32]Berthold K P Horn,Brian G Svhunck.Determing optic flow[J].Artificial Intelligence,1981:185-203.
[33]陈邦忠,王秀坛,涂建平,彭应宁.红外图象序列的运动目标检测。探测与控制学报,2002,24(3):12-20.
[34]田天,周兵,李波.基于解析小波的光流计算方法[J].北京航空航天大学学报,2003,29(6):548-551.
[35]Haussecker H W,Fleet D J.Computing optical flow with physical models of brightness variation[J].IEEE Trans Pattern Analysis and Machine Intelligence,2001,6(23):661-673.
[36]Wu Yute,Kanade T,Cohn J,et al.Optic flow estimation using wavelet motion model[A],computer Vision[c],1995:992-998.
[37]傅德胜,王海彬,孙文静.数字视频监控中的运动目标检测.微电子学与计算机,2005,2(1):118-212.
[38]Barron J,Fleet D,Beauchem in S.Performance of optical flow techniques[J].International Journal of Computer Vision,1994,12(1):42-77.
[39]Tretiak O,PastorL.Velocity Estimation from Image Sequence with Second Order Differential Operators[J].Proc.Of 7th ICPR,pp:16-19.
[40]LiptonA Fujiyoshi H,Patil R.Moving target classification and tracking from real-time video[J].Proc.Of WACV'98,1998:8-14.
[41]李超,熊章,赫阳,刘玉恒.基于帧间差的区域光流分析及其应用[J].计算机工程与应用,2005,31:195-198.
[42]孙承志,周军,吴明英,李新昊.运动目标实时跟踪系统的设计[J].河海大学常州分校学报,2006,20(1):28-31.
[43]Milan S,Vaclav H,Roger B.Image processing analysis and machine vision[M].2版.北京::人民邮电出版社,2002.
[44]OHTA Naoya.Optical flow detection by color images[J].NEC Research and Development,1990,97:78-84.
[45]OHTA Naoya.Uncertainty models of the gradient constraint for optical flow computation [J].IEICE Transactions on Information and Systems,1996,E79-D(7):958-964.
[46]贾云得.机器视觉[M].北京科学出版社,2000:235-239.
[47]万缨,韩毅,卢汉清.运动目标检测算法的探讨.计算机仿真,2006,23(10):221-225.
[48]Jianbo Shi,Carlo Tomasi.Good features to track[J].Proc.IEEE Comput,1994:593-600.
[49]Horn,BKP,BG Schunck.Determining optical flow[J].Artif Intell,1981,17:185-203.
[50]T Hirai,et al.Optical of small moving objects byoptical flow[C].In 11th International Conference on Pattern Reeognitio,Vol.Ⅱ,the Hague,Nehterlands,1992,pp:474-478.
[51]Firooz A Sadjadi,Cornell S L Chun.Automatic detection of small objects from their infrared state-of-polarization vectors[J].Optics Letters,2003,28(7),531-533.
[52]武理静,明军.一种改进的光流估计方法,国外电子测量技术,2006,25(12):6-8.
[53]陈伟,左敏.数字监控视频报警系统的设计[J].北京工商大学学报(自然科学版),2002,20(2):36-41.
[54]胡俊,苏祥芳.图像序列运动检测算法的研究及其应用[J].武汉大学学报(自然科学版),2000,46(5):613-616.
[55]胡以静,李政访,胡跃明.基于光流的运动分析理论及应用.计算机测量与控制, 2007,15(2):219-221.
[56]B Bhanu,P Symosek,S Snyder.Synergism of Binocular and Motion Stereo for Passive Ranging[J].IEEE,1994,30(3):523-532.
[2]赵天宇,吴易明,陈良益,朱立峰,刘宇波.结构光法和相位移法的对比研究[J].科学技术与工程[J],2007,7(17):4429-4432.
[3]赵勋杰,李成金.双目立体实时测距系统的关键技术研究[J].激光与红外,2006,36(9):874-877.
[4]金元柳.对激光近炸引信的干扰技术[J].光电技术应用,2004,19(5):7-10.
[5]赵易良.立体视觉与运动分析融合问题的研究[J].光电技术应用,2007:10-20.
[6]M.Bertozzi,A.Broggi,A.Fascioli.Stereo Vision-based vehicle detection[J].IEEE,2000,pp:39-44.
[7]吴利予,李金,吕普轶,李想.AVI文件的软件解析及自动单帧捕获[J].应用科技,2003,30(9):14-15.
[8]陶琳.基于光流技术的图像信息提取[J].华中科技大学学报,2006:30-55.
[9]J.L.Barron,D.J.Fleet,S.S.Beauchemin.Performance of Optical Flow Techniques[J].Computer Vision,1994,12(1):43-77.
[10]Milan Sonka,Vaclav Hlavac,Roger Boyle.Computer Vision[M].北京:人民邮电出版社[M],2006:466-477.
[11]陈兵旗,孙明.VC++使用图像处理专业教程[M].北京:清华大学出版社,2005:19-77.
[12]Ming Ye.Robust Visual Motion Analysis:Piecewise-Smooth Optical Flow and Motion-based Detection and Tracking[J].University of Washington,2002[0]
[13]Ronald J,Pieper,Ronald J,Pieper,G.Pelegris.Passive range estimation using dual-baseline Triangulation[J].Optical Engineering,1996,35(3):685-692.
[14]李耀辉,刘宝军.运动模糊图像的光流场分析[M].华北科技学院学报,2004,1(3):86-87.
[15]陈伏虎,罗雪松,陈洪兴,翁伟,费腾.双活动平台被动测距方法研究[J].声学与电子工程,2004,3(79):15-16.
[16]危水根,陈震,黎明.基于梯度光流法计算的一种改进[J].计算机工,2006,32(1):198-200.
[17]胡金华,吴文全.被动测距误差分析及修正[J].电子测量技术,2006,29(3):38-40.
[18]Yair Barniv.Error.Analysis of Combined Optical-Flow and Stereo Passive Ranging[J].Journal of the IEEE,1992,28(4):978-989.
[19]B.Sridhar,P.Smith,R.Suorsa,B.Hussien.Multirate and Event Driven Kalman Filters for Helicopter Passive Ranging[J].IEEE Control Systens Magazine,10(4):43-50.
[20]Yair Barniv.Passive Ranging Using Image Expansion[J].IEEE,1995,31(1):358-354.
[21]Maarten de Visser,Maarten de Visser,Maarten de Visser,Emile A.Hendriks.SPIE,2006,45(2):026402-1-14.
[22]陶琳,黄士科,张天序.基于光流的红外目标被动测距技术[J].计算机与数字工程,2005,33(11):1-4.
[23]黄士科,夏涛,张天序.基于红外图像的被动测距方法[J].红外与激光工程,2007,6(1):109-126.
[24]田旭,胡桂琴.AVI格式文件的分解处理[J],河北工业科技,2006,23(4):240-243.
[25]于晓康,柴乔林.基于P2nn的可控视频回放系统.计算机应用,2003,23:326-329.
[26]Inseok Hwang,Hamsa Balakrishnan,Kaushik Royt,and Claire Tomlin.Multiple-Target Tracking and Identity Management in Clutter,with Application to Aircraft Tracking.Proceeding of the 2004 American Control Conference,2004,1:3422-3428.
[27]Dr.Paul Frank Singer,Doreen M.Sasaki.Analysis of the Cascade of Track-Before-Detect and Track-After-Detect Tracking Algorithms[J].SPIE,1998,3373:157-165.
[28]Boris Rozovskii,Anton PetrovOptimal.Nonlinear filtering for track—before-detect in JR image sequences[J].SPIE,1999,3809:152-163.
[29]郭丰俊,杨新,W.Phliips,施鹏飞,I.Lemahieu.基于HUBER函数的序列图像位移场估计算法[J].高技术通讯,2001,34:38-42.
[30]白雪生,金国,徐光祜,史元春.基于帧间运动的视频分割[J].清华大学学报(自然科学版),2002,42(1):76-79。
[31]李熙莹,倪国强.红外图像的光流计算[J].红外与激光工程,2002,31(3):189-193.
[32]Berthold K P Horn,Brian G Svhunck.Determing optic flow[J].Artificial Intelligence,1981:185-203.
[33]陈邦忠,王秀坛,涂建平,彭应宁.红外图象序列的运动目标检测。探测与控制学报,2002,24(3):12-20.
[34]田天,周兵,李波.基于解析小波的光流计算方法[J].北京航空航天大学学报,2003,29(6):548-551.
[35]Haussecker H W,Fleet D J.Computing optical flow with physical models of brightness variation[J].IEEE Trans Pattern Analysis and Machine Intelligence,2001,6(23):661-673.
[36]Wu Yute,Kanade T,Cohn J,et al.Optic flow estimation using wavelet motion model[A],computer Vision[c],1995:992-998.
[37]傅德胜,王海彬,孙文静.数字视频监控中的运动目标检测.微电子学与计算机,2005,2(1):118-212.
[38]Barron J,Fleet D,Beauchem in S.Performance of optical flow techniques[J].International Journal of Computer Vision,1994,12(1):42-77.
[39]Tretiak O,PastorL.Velocity Estimation from Image Sequence with Second Order Differential Operators[J].Proc.Of 7th ICPR,pp:16-19.
[40]LiptonA Fujiyoshi H,Patil R.Moving target classification and tracking from real-time video[J].Proc.Of WACV'98,1998:8-14.
[41]李超,熊章,赫阳,刘玉恒.基于帧间差的区域光流分析及其应用[J].计算机工程与应用,2005,31:195-198.
[42]孙承志,周军,吴明英,李新昊.运动目标实时跟踪系统的设计[J].河海大学常州分校学报,2006,20(1):28-31.
[43]Milan S,Vaclav H,Roger B.Image processing analysis and machine vision[M].2版.北京::人民邮电出版社,2002.
[44]OHTA Naoya.Optical flow detection by color images[J].NEC Research and Development,1990,97:78-84.
[45]OHTA Naoya.Uncertainty models of the gradient constraint for optical flow computation [J].IEICE Transactions on Information and Systems,1996,E79-D(7):958-964.
[46]贾云得.机器视觉[M].北京科学出版社,2000:235-239.
[47]万缨,韩毅,卢汉清.运动目标检测算法的探讨.计算机仿真,2006,23(10):221-225.
[48]Jianbo Shi,Carlo Tomasi.Good features to track[J].Proc.IEEE Comput,1994:593-600.
[49]Horn,BKP,BG Schunck.Determining optical flow[J].Artif Intell,1981,17:185-203.
[50]T Hirai,et al.Optical of small moving objects byoptical flow[C].In 11th International Conference on Pattern Reeognitio,Vol.Ⅱ,the Hague,Nehterlands,1992,pp:474-478.
[51]Firooz A Sadjadi,Cornell S L Chun.Automatic detection of small objects from their infrared state-of-polarization vectors[J].Optics Letters,2003,28(7),531-533.
[52]武理静,明军.一种改进的光流估计方法,国外电子测量技术,2006,25(12):6-8.
[53]陈伟,左敏.数字监控视频报警系统的设计[J].北京工商大学学报(自然科学版),2002,20(2):36-41.
[54]胡俊,苏祥芳.图像序列运动检测算法的研究及其应用[J].武汉大学学报(自然科学版),2000,46(5):613-616.
[55]胡以静,李政访,胡跃明.基于光流的运动分析理论及应用.计算机测量与控制, 2007,15(2):219-221.
[56]B Bhanu,P Symosek,S Snyder.Synergism of Binocular and Motion Stereo for Passive Ranging[J].IEEE,1994,30(3):523-532.