摘要
现代靶场为适应武器系统快速发展的需要，提出了对小、暗、多、快目标测量的新要求。由于目标具有小、多、不发光和速度快等特点，要获取目标的清晰图像，相应的光学测量设备必须具有高帧频、高灵敏度和大视场的特性。基于对成本和安全性的要求，光学测量设备的布站位置距目标相对较远，因此无法给出对小、暗、多、快目标作用过程的细节。本文结合现代电子通信技术与摄像测量方法从低成本高速CMOS摄像机设计、图像采集存储、摄像机标定和目标测量等关键技术对该问题进行了深入的研究。本文的研究内容主要集中在以下几个方面：
（1）根据测量要求并考虑研制成本，本文设计了一款分辨率为2K×1K、最高帧频为280帧、曝光可调的低成本高速CMOS摄像机，并详细说明了高速摄像机设计中的关键技术，尤其对高速数据采集中的位对齐和字对齐技术做了重点论述。为了适应测量设备小型化、便携化和低功耗的发展趋势，本文设计了一套基于高速SATA传输接口的嵌入式图像存储系统。实验结果表明，该图像存储系统最高图像存储速度可以达到294MB/s。
（2）摄像机标定算法在本文中做了详细讨论和深入研究。以平面棋盘为标定工具，采用平面标定算法标定测量系统摄像机的内外参数，并针对该算法在实际标定中标定结果不收敛的问题，在随机抽样一致性（RANSAC）算法的基础上，提出了一种改进的标定算法。实验结果表明，改进的标定算法与传统方法相比，在降低算法迭代次数，优化算法运行效率的同时，标定精度提高了35%。
（3）根据小、暗、多、快目标测量的实际需求，本文采用平面摄像测量方法单目测量目标的脱靶量和落点坐标，该方法具有测量简便、灵活性高的优点。在此基础上，针对点目标抛物线运动测量的问题，证明了二次抛物线的射影几何性质，并提出了一种针对抛物线运动目标三维信息测量的单目测量方法。通过仿真和实际小球投掷实验，验证了方法的正确性和可行性。该方法与双目测量结果比较，小球三维坐标测量误差在1.2%以内。
（4）针对更一般的交汇测量任务，本文首先详细介绍了交汇测量中的关键技术，包括三维欧式重构、立体匹配与外参数标定等；其次，重点分析了外参数对测量精度的影响，通过仿真得到了最优的结构参数；最后，针对小、暗、多、快目标的落点测量问题，提出了一种无需标定的平面交汇测量方法，该方法可以应用于大范围、远距离、粗测量的场合。
With the rapid development of weapons systems, new requirements are raisedfor measurement of small, dark, rapid, multi targets in modern shooting ranges.Since the rapid, multi targets are smalland non-luminous, to get clear pictures of thetargets, corresponding optical measurement equipment must have the characteristicsof high frame rate, high sensitivity and large field of view. For reasons of safety andcost, optical measurement equipments are posited far from the targets, thereforedetails of the targets movement are missed. Based on modern electroniccommunication technology and camera measurement technique, image acquisitionand storage, camera calibration and targets measurement are taken into considerationthis problem in this dissertation. The main works are summed in follows:
(1) A new low-cost high speed camera is designed with the consideration ofmeasurement requirements and hardware costs. The image resolution is2Kx1K andthe maximum frame frequency is up to280fps under the condition of full resolution.Exposure of the camera can adjust to application environment. The key technologiesof camera design are explained in details, especially for bit alignment and wordalignment technique in the process of high-speed data acquisition. Under theminiaturization of the measuring device, trend of better portability and low powerconsumption, an embedded image storage system is also designed with thehigh-speed SATA interface. The experimental results indicate that the bandwidth of storage system can reach294MB/s。
(2) In this paper, the camera calibration technology is in-depth researched. Theexternal and internal parameters of camera are estimated using the plane calibrationalgorithm with planar checkerboard. The plane calibration algorithm is easy andflexible, but sometimes the algorithm is not convergent in practice when the numberof calibration imagesis more than10. A modified method is proposed to ensurecalibration convergence. The modified method divides the calibration images intocorrect images and false images using Random Sample Consensus（RANSAC）algorithm, and calibrates camera only for good images. The experimental resultsshow that the robust modified method can reduce the number of iteration, andcalibration precision increase by35%.
(3) To satisfy actual demand for the measurement of small, dark, rapid, multitargets, planar imaging measurement, easy and high flexibly, is used to monocularmeasure the miss distance and drop point coordinates of targets in this paper. On thisbasis, nature of the projective geometry of parabola is proved and a novel monocularmeasurement method is proposed for parabolic motion of point targets. Bothcomputer simulation and real data have been used to verify the feasibility of thismethod.. Practical globule throwing experiments show that the measurement erroris less than1.2%compared with binocular measurement methods.
(4) For general measurement tasks, the key technologies for intersectionmeasuring are firstly introduced, which include3D European reconstruction, stereomatching and external parameters calibration, etc. Secondly, the influence ofexternal structure parameters on the accuracy of measurement is analyzedemphatically, and the optimal parameters are achieved by numerical simulation.Finally, Aiming at measurement of the small, dark, rapid, multi targets drop point, anovel plane intersection method is raised and it can be applied in large view field,long distance and coarse measurement occasions.
引文
[1]张广军.视觉测量[M].北京:科学出版社,2008.
[2]迟健男.视觉测量技术[M].北京：机械工业出版社，2011.
[3]邾继贵，于之靖.视觉测量原理与方法[M].北京：机械工业出版社，2012.
[4]冈萨雷斯.数字图像处理[M].北京：电子工业出版社，2004.
[5]邸慧.靶场图像预处理和目标运动单像机测量方法研究[D]:[博士学位论文].长沙：国防科技大学,2007.
[6] Rosin P L, Ellis T. Image difference threshold strategies and shadowdetection[C].6th British Machine Vision Conf., Birmingham,1995,347-356.
[7] Mae Y, Shirai Y, Miura J, et al. Object tracking in cluttered background based onoptical flow and edges[C]. Proceedings of the13th International Conference onPattern Recognition,1996,1:196-200.
[8] Araki Shoichi, Matsuoaka Takashi, Yokoya Naokazu, Takemura Harua. Real-timetracking of multiple moving object contours in a moving camera image sequence[J].IEICE Trans. Inf.&Syst.,2000, E83-D(7):131-137.
[9]李红艳.图像低信噪比小目标检测与跟踪算法研究[D]:[博士学位论文]，西安：西安电子科技大学，2000.
[10]张淑芬，李华.基于一幅散焦图像的深度估计新方法[J].光电子激光[J],2006,17(3):364-367.
[11]于起峰,孙祥一,权铁汉等.用标定和亚像素技术实现三维运动目标的高精度测量[J].宇航学报，1999,20(3):38-42.
[12]于起峰，尚洋.摄像测量学原理与应用研究[M].北京：科学出版社，2009.27-32,244-250
[13]黄凤山，钱惠芬.光笔式单摄像机三维坐标测量视觉系统[J].光电工程，2007,34（4）：69-72.
[14]贾小军,张之江,曹芳等.编码结构光系统模型及误差分析[J].光学精密工程,2011,19(4):717-726.
[15]张淑芬，李华.基于一幅散焦图像的深度估计新方法[J].光电子激光[J],2006,17(3):364-367.
[16] Lowe.Distinctive Image Features from Scale-Invariant Keypoints[J].International Journal of Computer Vision,2004,60(2):91-110.
[17] Krystian Mikolajczyk,Cordelia Schmid. Scale&Affine Invariant Interest PointDetectors[J].International Journal of Computer Vision,2004,60(1):63-86.
[18]丁南南.基于特征点的图像配准技术研究[D]:[博士学位论文].长春：中国科学院长春光学精密机械与物理研究所，2012.
[19]王晓欧.基于扩展卡尔曼滤波的摄像机标定方法研究[D]:[硕士学位论文].哈尔滨:哈尔滨工业大学，2008.
[20] Daniel Scharstein, Richard Szeliski.A Taxonomy and Evaluation of DenseTwo-Frame Stereo Correspondence Algorithms[J]. International Journal of ComputerVision,2002,47(1/2/3):7-42.
[21] Guanghui Wang,Zhanyi Hu,Fuchao Wu,et al. Single view metrology from sceneconstraints[J]. Image and Vision Computing23(2005)831–840.
[22] Guanghui Wang,Hung-Tat Ysui,Zhanyi Hu,,et al. Camera calibration and3Dreconstruction from a single view based on scene constraints[J]. Image and VisionComputing23(2005)311–323.
[23] R. Y. Tsai.A versatile camera calibration technique for high accuracy3D machinevision metrology using off-the-shelf TV cameras and lenses[J].IEEE J. RoboticsAutomat,1987,3(4):323-344.
[24]尚洋.基于视觉的空间目标位置姿态测量方法研究[D]:[博士学位论文]，长沙：国防科技大学，2006.
[25] Hartley, R.I., Zisserman, A.. Multiple View Geometry in Computer Vision,second edtion[M], Cambridge University Press,2004.
[26]吴福朝.计算机视觉中的数学方法[M].北京:科学出版社,2011.
[27] Richard Szeliski.Computer Vision: Algorithms and Application [M].German:Springer Press,2010:41-46.
[28] Weng J, Cohen P, Herniou M. Camera calibration with distortion models andaccuracy evaluation[J].IEEE Transactions on Pattern Analysis and MachineIntelligence,1992,14(10):965-980.
[29] Salvi J, Armangue X, Batle J. A comparative review of camera calibratingmethods with accuracy evaluation [J].Pattern Recognition,35(2002)1617-1635.
[30]马颂德,张正友.计算机视觉-计算理论与算法基础[M].北京:科学出版社,1998.
[31] Jean-Yves Bouguet.Camera calibration Toolbox for Matlab[EB/OL].http://www.vision.caltech.edu/bouguetj/calib_doc/.
[32] Zhengyou Zhang. A flexible new technique for camera calibration[J].IEEETrans.Pattern Analysis and Machine Intelligence2000,22(11):1330-1334.
[33] Peter Kovesi. MATLAB and Octave Functions for Computer Vision and ImageProcessing[EB/OL], www.csse.uwa.edu.au/~pk/research/matlabfns/.
[34]袁亚湘，孙文瑜.最优化理论与方法[M].北京：科学出版社，2001.
[35]周国辉.CCD摄影测量相机图像数据高速实时存储的研究[D]:[博士学位论文].长春:中国科学院长春光学精密机械与物理研究所,2006.
[36]王桂强.基于IEEE1394总线的高速相机数据传输方案设计[D].哈尔滨：哈尔滨工程大学，2011.
[37]吕耀文，王建立，曹景太.Base型Camera Link脱机存储系统设计[J].光电子技术，2012,32（4）：242-245.
[38] Automated Imaging Association. Camera Link Specification[EB/OL]. Revision1.0, January2004.
[39] CMOSIS Incorporation.2.2Megapixel Machine Vision CMOS image sensordatasheet [EB/OL].Revision1.0, November2001.
[40]宋亚军,许廷发,倪国强,等.基于Virtex-4FPGA的低功耗图像融合系统[J].光学精密工程,2007,15(6):935-940.
[41]刘彬,房建成,刘刚.基于TMS320C6713B+FPGA数字控制器实现磁悬浮飞轮主动振动控制[J].光学精密工程,2009,17(1):151-157.
[42] Xilinx Inc. Virtex-4FPGA Data Sheet: DC and Switching Characteristic[EB/OL].[2009-09-09]. http://www.xilinx.com.
[43] Abhijit Athavale, Carl Christensen.High-Speed Serial I/O Made Simple [M],Xilinx Inc，2005.
[44] Technical Committee of Accredited Standards Committee.InformationTechnology AT attachment with packet interface-7[EB/OL].Revision1.0, November2001.
[45] InnoDisk Inc. Embedded DiskCard8000DataBook[EB/OL].[2012-02-18].http://www.innodisk.com.
[46]徐启明,张启衡,张耀.基于硬盘直写的脱机高速图像记录系统[J].传感器与仪器仪表,2009,25(1):120-122.
[47]维基百科.SATA[EB/OL]. http://zh.wikipedia.org/wiki/SATA.
[48]范俊.基于FPGA的SATA控制器的研究与实现[D]:[硕士学位论文].武汉：华中科技大学，2006.
[49]龙小奇.基于SATA接口的固态硬盘控制器设计[D]:[硕士学位论文].长沙：湖南大学，2009.
[50]周双喜.基于FPGA的高速数据记录系统的研究[D]:[硕士学位论文].成都：电子科技大学，2010.
[51]叶项胜.基于FPGA的Serial ATA1.0设备IP CORE设计[D]:[硕士学位论文].成都：西南石油大学，2006.
[52] Xilinx Inc.ML505/ML506/ML507Evaluation Platform User Guide [EB/OL],Revision3.0www.xilinx.com.2008-5-19.
[53] Xilinx. LightWeight IP Application Examples[EB/OL], Revision3.1,www.xilinx.com.2011-4-21.
[54] Xilinx. LogiCore IP Multi-Port memory Controller(MPMC)[EB/OL],Revision6.0.3a, www.xilinx.com.2011-3-1.
[55]叶肇晋,张稀楠,马磊.基于Xilinx FPGA片上嵌入式系统的用户IP开发[M].西安：西安电子科技大学出版社，2008.
[56]郑志刚.高精度摄像机标定和鲁棒立体匹配算法研究[D]:[博士学位论文].合肥：中国科学技术大学，2008.
[57]田震.基于神经网络方法的摄像机标定技术研究[D]:[硕士学位论文].长沙：国防科学技术大学，2010.
[58]瞿静，孙瑞，童癸.计算机视觉中的摄像机标定技术综述[J].电子技术，2011.
[59]陈铎.摄像机标定方法研究与实现[D]:[硕士学位论文].大连：东北大学，2009.
[60] De Xu,You Fu Li,Yang Shen,et al.New Pose-Detection Method forSelf-Calibrated Cameras Based on Parallel Lines and Its Application in Visual ControlSystem[J]. IEEE TRANSACTIONS ON SYSTEMS, MAN, ANDCYBERNETICS—PART B: CYBERNETICS.2006,36(5):1104-1107.
[61] Zhaoxiang Zhang,Tieniu Tan,Kaiqi Huang,et al. Practical Camera CalibrationFrom Moving Objects for Traffic Scene Surveillance[J]. IEEE TRANSACTIONS ONCIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY,2013,23(3):518-533.
[62]黄风山,钱惠芬.三坐标测量机驱动的摄像机标定技术[J].光学精密工程,2010,18(4):952-967.
[63]杨博文，张丽艳，叶南等.面向大视场视觉测量的摄像机标定技术[J].光学学报，2012，32(9):0915001.
[64]牛海涛,赵勋杰.采用棋盘格模板的摄像机标定新方法[J],红外与激光工程,2011,40(1):1342-1346.133-137.
[65] Fei Qi,Qihe Li,Yupin Luo,et al. Constraints on general motions for cameracalibration with one-dimensional objects[J] Pattern Recognition40(2007)1785–1792.
[66] Zhengyou Zhang. Camera Calibration with One-Dimensional Objects[J] IEEETransactions on Pattern Analysis and Machine Intelligence,2004,26(7):892-899.
[67] Xiaoqiao Meng,Zhanyi Hu, A new easy camera calibration technique based oncircular points[J]Pattern Recognition36(2003)1155-1164.
[68] S.J. Maybank and O.D. Faugeras, A Theory of Self-Calibration of a MovingCamera[J].International Journal of Computer Vision,1992,8(2),123-152.
[69] Sterm P.A case against Kruppa’s equations for camera self-calibration [C].InProceedings of International Conference of Image Processing, Illinois,1998,172-175.
[70]张虎，达飞鹏，李勤.基于两相同圆的自标定方法[J].仪器仪表学报，2010，31(3):618-624.
[71]吴福朝，王光辉，胡占义.由矩形确定摄像机内参数与位置的线性方法[J].软件学报，2003，14（3）：703-712.
[72] Aruro de la, Jose Maria. Automatic Chessboard Detection for Intrinsic andExtrinsic Camera Parameter Calibration[J].Sensors,2010,10,2027-2044.
[73] Ha,J.E. Automatic detection of chessboard and its applications. Opt. Eng.2009,48,067205.
[74] Institute of Robotics and Mechatronics. DLR Camera CalibrationToolbox[EB/OL]. www.dlr.de/rm/en/desktopdefault.aspx/tabid-3925/.
[75]于仕琪,刘瑞帧.学习OpenCV[M].北京：清华大学出版社,2009.
[76] Martin A.Fischler, Robert C.Bolles. Random Sample Consensus: A Paradigm forModel Fitting with Appicatlons to Image Analysis and Automated Cartography[J].Communications of the ACM,1981,24(6):381-395.
[77] Ankur Datta,Jun-Sik Kim,Takeo Kanade. Accurate Camera Calibration usingIterative Refinement of Control Points[C]. IEEE12th International Conference onComputer Vision Workshops,1999,1201-1208.
[78] Fishler, M. A., and R. C. Bolles. Random sample consensus: A Paradigm formodel fitting with applications to image analysis and automatedcartography[J].Communications of the ACM,1981,24(6):381-395.
[79] Daniel F.Dementhon and Larry S.Davis. Model-Based Object Pose in25lines ofcode[J]. International Journal of Computer Vision,1995,15(2):123-141.
[80]吴福朝，胡占义. PnP问题的线性求解算法[J].软件学报，2003,14(3):682-688.
[81]周维，许海霞.一种摄像机姿态估计方法[J].系统仿真学报，2011，23(12)：2582-2587.
[82] X.S. Gao,J.Tang,On the probability of the number of solutions forthe P4P problem[J], Math. Imaging Vision,2006,25(1)79–86.
[83] Z.Y. Hu,F.C.Wu, A note on the number of solutions of the noncoplanar P4Pproblem,IEEE Trans.Pattern Anal. Mach. Intell.24(4)(2002)550–555.
[84]周鑫,朱枫.关于P3P问题解的唯一性条件的几点讨论[J].计算机学报,2003,26(12):1696-1701.
[85]胡占义，雷成，吴福朝.关于P4P问题的一点讨论[J].自动化学报，2001,27（6）：770-776.
[86]吴福朝，胡占义.关于P5P问题的研究[J].软件学报，2001,12(5):768-775.
[87]邾继贵，李艳军，叶声华等.单摄像机虚拟立体视觉测量技术研究[J].光学学报，2005,25（7）：943-948.
[88]宋振东，姜伟.单目多视点立体图像提取与应用[J].光学学报，2012，32(5)：0515001.
[89] NaoKi YAMAGUTI, Shunichiro OE, Kenji TERADA. A Method of DistanceMeasurement by Using Monocular Camera[C].SICE’97. Proceedings of the36thSICE Annual Conference, International Sesion Papers,1997:1255-1260.
[90] Liu Xiaoming, Qin Tian, Chen Wanchun. Real-Time Distance MeasurementUsing a Modified Camera[C] IEEE.sensors applications symposium(SAS),2010:54-58.
[91]韩延祥，张志胜，戴敏.用于目标测距的单目视觉测量方法[J].光学精密工程，2011,19(5):1110-1116.
[92]霍炬，杨卫，杨明.基于消隐点几何特征的摄像机自标定方法[J].光学学报，2010，30(2)：465-472.
[93] Guanghui Wang, Jonathan Wu, Zhengqiao Ji, Single View based pose estimationfrom circle or parallel lines[J].Pattern Recognition Letters,29(2008):977-985.
[94]唐正宗，梁晋，肖振中等.用于三维变形测量的数字图像相关系统[J].光学精密工程，2010，18（10）：2244-2253.
[95]金伟伟.基于双目视觉的运动小目标三维测量的研究与实现[D]:[硕士学位论文].杭州：浙江大学，2010.
[96]宋晓伟.双目视觉位姿测量算法研究[D]:[硕士学位论文].北京：中国地质大学，2011.
[97]侯俊捷，魏新国，孙军华.基于同心圆合成图像匹配的双目视觉标定[J].光学学报，2012，32(3)：0315003-1.
[98]张琼，王志良，迟健男等.基于平面镜面面向双摄像机视线追踪系统的标定方法[J].光学学报，2011，31(4)：0415002.
[99] Kenichi Kanatani, Yasuyuki Sugaya, Hirotaka Niitsuma. Triangulation from TwoViews Revisited: Hartly-Sturm vs. Optimal Correction[C],19th British MachineVision Conference(BMVC2008),Leeds U.K:2008:173-182
[100] Torr P H S,Murray D W. The development and Comparison of Robust Methedsfor Estimating the Fundamental Matrix[J].International Journal of ComputerVision,1997,24(3):1-33.
[101] J.David, D.Allan, et al.Phase-based disparity measurement[J].Imageunderstanding archive.1991,53(2):198~210.
[102]朱庆生,胡章平,刘然等.立体图像对的极线校正[J].计算机工程与技术，2009,30（7）：4027-4030.
[103]曾吉勇,苏显渝.一种无相机标定的立体图像对校正新方法[J].光学学报，2004,24(5):628-632.
[104]罗桂娥.双目立体视觉深度感知与三维重建若干问题研究[D]:[博士学位论文],长沙：中南大学，2012.
[105]]弥丽丽.基于双视觉的复杂背景下运动目标三维信息重建[D]:[硕士学位论文]，西安：西北大学：2010.
[106]夏永泉.计算机视觉中双目匹配相关技术研究[D]:[博士学位论文],南京：南京理工大学：2007.
[107]董圣男.基于双目立体视觉的空间非合作目标的位姿测量[D]:[硕士学位论文]，南京：南京航空航天大学，2010.
[108]池凌鸿.立体匹配算法的研究和应用[D]:[博士学位论文],合肥：中国科学技术大学，2010.
[109]刘俸材，谢明红，颜国霖.双目立体视觉系统的精度分析[J].计算机工程，2011,37（19）：280-282.
[110]闫龙,赵正旭,周以齐.基于CCD的立体视觉测量系统精度分析与结构设计研究[J].仪器仪表学报,2008,29(2):410-413.
[111]陈泽志，吴成柯，沈沛意.计算机视觉测量系统的误差模型分析[J].计算机辅助设计与图形学学报,2002,14(5):389-393.
[112]刘琼,秦现生，应申舜等.双目视觉测量系统结构参数设计及精度分析[J].中国机械工程,2008,19(22):2728-2732.