低信噪比图像中多运动目标的实时识别研究与应用
|
摘要
本文首先概述了模式识别的分类过程、系统设计和应用领域，分析了低信噪比图像中多运动目标实时识别的难点，重点介绍了本课题的研究内容以及解决方案。
接着，讲述了分割图像的三种结构和传统的基于边缘、边界和区域的图像分割技术以及几种常见的颜色模型。然后，分析和阐明了传统的运动目标检测方法的不足，并在此基础上结合研究中的实际实验环境，提出了一系列解决方法，包括针对降低庞大数据量而提出的网格扫描、局部“跟虫”追踪和动态窗口扫描等目标检测方法，针对实验环境中光照不均和颜色干扰提出基于人机交互的最大最小值阈值选取方法和引入改进的RGB模型到HSV模型的转换方法，为消除图像畸变对识别精度的恶劣影响而采用的通过控制点进行双线性插值进行畸变校正的方法；
紧接着，概述了神经网络的发展历史和几种常用神经网络模型的特点，重点研究了前馈型神经网络在模式识别中的应用问题，详细阐述了基本的BP算法和学习过程中BP算法的改进，从而使网络收敛速度更快，解决问题更有效，并在此基础上，设计了一个基于BP神经网络的运动目标识别系统，给出了实验结果。
机器人足球系统作为本课题研究的一个应用实例，本文在最后，介绍了机器人足球比赛的学术背景、战略意义和在现实生活以及军事上的应用前景，阐明了机器人足球比赛系统的三种控制结构，由此指出半自主型机器人足球系统是一个合适的研究平台，并分析了这种系统各个组成部分的工作原理和功能，深入讨论了整个视觉软件的实现过程，并系统地阐述了在Windows系统平台下进行图像采集卡的二次开发方法，对从事实时图像处理、图像识别和硬件二次开发的人员有一定的借鉴作用。
Firstly, the process of classifying, system design and application of pattern recognition are summarized in this thesis. The difficulties in recognizing several moving objects in real-time images with low SNR are analyzed. And then the research contents as well as resolution are both explained with emphases.
Next, the three saving methods for image segmentation results, the traditional technologies of image segmentation based on edge, boundary and region, together with several common color spaces are introduced. Later on, after elaborating the disadvantages of the old methods in detecting and recognizing moving objects, a series of corresponding approaches are proposed, such as grid scan, local tracking bug and dynamic window in object tracing to reduce the huge data needed to be processed, maximum and minimum for selecting a proper segmentation threshold and improved conversion from RGB model to HSV and so on to decrease the influence of inhomogeneous lighting and the color noise, a bilinear interpolation in each quadrant to eliminate the bad effect on the recognition precise because of the distortions of the camera.
After that, much emphasis is given on application study in pattern recognition with a feed-forward neural network. Both the basic BP algorithm and improved BP algorithm in the study process are described in detail, and the later is used to quicken convergence speed and improve validity of the network. According to the above-mentioned introduction, the author has designed a recognition system of moving objects based on BP neural network and experimental results are showed.
Finally, as RoboCup is an applied example of this project, the learning background, strategy significance and the applied foreground in our daily life and military affairs of the robotic soccer system are summarized in this thesis. After introducing of the three control constructions in robotic soccer system and comparing with each other, we pointed out the half-independent robotic soccer system is more appropriate to be our
study platform. The analysis of the principle and function of each component in this kind of system and the whole process of how to carry out the vision software system from the input image to the information of robots and ball and the ways in a second development of the frame grabber on the Windows system platform are both perfectly described. This will make a great help to those who do a study in real-time image processing.
接着，讲述了分割图像的三种结构和传统的基于边缘、边界和区域的图像分割技术以及几种常见的颜色模型。然后，分析和阐明了传统的运动目标检测方法的不足，并在此基础上结合研究中的实际实验环境，提出了一系列解决方法，包括针对降低庞大数据量而提出的网格扫描、局部“跟虫”追踪和动态窗口扫描等目标检测方法，针对实验环境中光照不均和颜色干扰提出基于人机交互的最大最小值阈值选取方法和引入改进的RGB模型到HSV模型的转换方法，为消除图像畸变对识别精度的恶劣影响而采用的通过控制点进行双线性插值进行畸变校正的方法；
紧接着，概述了神经网络的发展历史和几种常用神经网络模型的特点，重点研究了前馈型神经网络在模式识别中的应用问题，详细阐述了基本的BP算法和学习过程中BP算法的改进，从而使网络收敛速度更快，解决问题更有效，并在此基础上，设计了一个基于BP神经网络的运动目标识别系统，给出了实验结果。
机器人足球系统作为本课题研究的一个应用实例，本文在最后，介绍了机器人足球比赛的学术背景、战略意义和在现实生活以及军事上的应用前景，阐明了机器人足球比赛系统的三种控制结构，由此指出半自主型机器人足球系统是一个合适的研究平台，并分析了这种系统各个组成部分的工作原理和功能，深入讨论了整个视觉软件的实现过程，并系统地阐述了在Windows系统平台下进行图像采集卡的二次开发方法，对从事实时图像处理、图像识别和硬件二次开发的人员有一定的借鉴作用。
Firstly, the process of classifying, system design and application of pattern recognition are summarized in this thesis. The difficulties in recognizing several moving objects in real-time images with low SNR are analyzed. And then the research contents as well as resolution are both explained with emphases.
Next, the three saving methods for image segmentation results, the traditional technologies of image segmentation based on edge, boundary and region, together with several common color spaces are introduced. Later on, after elaborating the disadvantages of the old methods in detecting and recognizing moving objects, a series of corresponding approaches are proposed, such as grid scan, local tracking bug and dynamic window in object tracing to reduce the huge data needed to be processed, maximum and minimum for selecting a proper segmentation threshold and improved conversion from RGB model to HSV and so on to decrease the influence of inhomogeneous lighting and the color noise, a bilinear interpolation in each quadrant to eliminate the bad effect on the recognition precise because of the distortions of the camera.
After that, much emphasis is given on application study in pattern recognition with a feed-forward neural network. Both the basic BP algorithm and improved BP algorithm in the study process are described in detail, and the later is used to quicken convergence speed and improve validity of the network. According to the above-mentioned introduction, the author has designed a recognition system of moving objects based on BP neural network and experimental results are showed.
Finally, as RoboCup is an applied example of this project, the learning background, strategy significance and the applied foreground in our daily life and military affairs of the robotic soccer system are summarized in this thesis. After introducing of the three control constructions in robotic soccer system and comparing with each other, we pointed out the half-independent robotic soccer system is more appropriate to be our
study platform. The analysis of the principle and function of each component in this kind of system and the whole process of how to carry out the vision software system from the input image to the information of robots and ball and the ways in a second development of the frame grabber on the Windows system platform are both perfectly described. This will make a great help to those who do a study in real-time image processing.
引文
[1] 郑南宁．计算机视觉与模式识别．北京：国防工业出版社，1998
[2] 蔡元龙．模式识别．西安：西北电信工程学院出版社，1986
[3] 朱志刚．数字图像处理．电子工业出版社，1998
[4] 戚飞虎．模式识别与图像处理．上海：上海交通大学出版社，1989
[5] 付京孙．模式识别及其应用．北京：科学出版社，1983
[6] Kenneth R. Castleman. Digital Image Processing. Prentice-Hall, 1996
[7] 胡守仁．神经网络导论．长沙：国防科技大学出版社，1993
[8] 贾云得．机器视觉．北京：科学出版社，2000
[9] Piccioli G, MicheliD, Parodi P, et al. A robust method for road sign detection and recognition. Image and Vision Computing, 1996, 14 (1): 209-233
[10] P. Nordlund, T. Uhlin. Closing the loop: detection and pursuit of a moving observer. Image and Vision Computing, 1996, 14(1):265-275
[11] Clarke J C, Zisserman A. Detection and tracking of independent motion. Image and Vision Computing, 1996, 14(1):565-572
[12] Vieren C, Cabestaing F, Postaire J. Catching moving objects with snakes for motion tracking. Pattern Recognition Letters, 1995, 16(1): 679-685
[13] Hogg D. Model-based vision: A program to see a walking person. Image and Vision Computing, 1995, 1(1):5-20
[14] Rowe S, Blake A. Statistical mosaics for tracking. Image Vision Computing, 1995, 14(1):549-564
[15] Jolly M D. Vehicle Segmentation and Classification Using Deformable Templates. IEEE TRANS. PAMI, 1996, 18(3):225-234.
[16] 杨杨，张田文．基于多分辨率方法的主动轮廓线跟踪算法．计算机学报，1998，21(3)：210-216
[17] 吴一全，朱兆达．图象处理中阈值选取方法30年(1962-1992)的进展(一)．数据采集与处理，1993，8(3)：193-201
[18] 吴一全，朱兆达．图象处理中阈值选取方法30年(1962-1992)的进展(二)．数据采集与处理，1993，8(4)：268-277
[19] 余新平，朱立．一种具有抗噪声干扰的图象边缘提取算法的研究．电子技术应用，1999，25(1)：9-10
[20] 黄伟．人脸识别及其应用研究：[硕士学位论文]．长沙：湖南大学，2001
[21] 彭涤．限制性手写体汉字特征点、笔划段特征的抽取算法研究：[硕士学位论文]．长沙：湖南大学，1993
[22] 伯晓晨，李涛，刘路，等．Matlab工具箱应用指南——信息工程篇．电子工业出版社，2000
[23] 赵景秀，王刚．基于彩色信息区分度的彩色边缘检测．计算机应用，2001，
[24] 姚鸿勋．基于彩色图像的色系坐标变换的面部定位与跟踪法．计算机学报，2002；2(1)：32-35
[25] Casasent D, Linlin Y, Slaski J. Opticalmuti-target subpixel detection and tracking initiation. SPIE, Optical Computing and Nonlinear Materials. 1988, 8(8):12-26
[26] Verri A, Poggio T. Against quantitative optical flow. Proc. of First Conference on Computer Vision. London, 1987
[27] Shahriar Negahdaripout. Revised definition of optical flow: integration of radiometric and geometric cues for dynamic scene analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998, 20(9):961-979
[28] Barron J L, Fleet D J, Beauchemin S S. Performance of optical flow techniques. International Journal of Computer Vision, 1994,12(1):43-47
[29] 张杏，彭晓燕，谭鹤良．足球机器人系统中一种快速模式识别算法的研究与实现．湖南大学学报，2002，6(11)：33-37
[30] Giovanni Adorni, Stefano Cagnoni. Vision-based localization for mobile robots. Robotics and Autonomous Systems, 2001(36):103-119
[31] 佟国峰，赵姝颖，徐心和．提高足球机器人彩色视觉系统实时性的若干图像识别关键技术研究．中南工业大学学报，2000，10(11)：497-500
[32] 欧宗瑛，袁野，张艳珍．基于颜色信息足球机器人视觉跟踪算法．大连理工大学学报，2000，40(6)：729-732
[33] Andrea Bonarini, Paolo Aliverti, Michele Lucioni. An Omni directional Vision Sensor for Fast Tracking for Mobile Robots. IEEE Trans. Instrumentation and Measurement, 2000(49): 509-512
[34] 杨金钟，刘政凯．基于控制点的图象变形方法及其应用．中国图象图形学报，2001，11(6)：1070-1074
[35] Haralick R.M., Shapiro L. G.. Computer and Robot Vision. Mass: Addision-Weslay Publishing Company, 1992
[36] 张勇．机器人足球识别算法与决策仿真：[硕士学位论文]．长沙：湖南大学，2002
[37] R. C. Eberhart, R. W. Dobbins. Neural Network Pe tools. Academic Press,1990
[38] 勒蕃，范俊波．神经网络与神经计算机．西南交通大学，1991
[39] Teuvo kohonen. The Self-organization Map. Proceedings of the IEEE,1990
[40] D.G. Bounds, P.J. Flioyd, B. Matlew, et al. A Multilayer Perceptron Network for the Diagnosis of Low Back Pain. Proc. 2d IEEE Int. Conf. on Neural Networks, London, 1988
[41] P.M. Shea, V. Lin.. Detetical of Explosives in Checked Airline Baggage Using an Artificial Neural Systems. Proc. Int. Joint Conf. on Neural
works, London, 1989
[42] T.J. Sejnowski, C.R. Rosenberg. Parallel Networks that Learn to Pronounce English Text. Complex Systems, 1987
[43] D. Nguyen, B. Widrow. The Truck Backer-upper: An Example of Self-learning in Neural Networks. Proc. Int. Joint Conf. on Neural Networks, Londan, 1989
[44] 庄镇泉，王煦法，王东生．神经网络与神经计算机．北京：科学出版社，1992
[45] Bernard Widrow, Micheal A. Lehr. 30 Years of Adaptive Neural Networks:Perceptron, Madline and Backpropagation. Proc. of the IEEE, 1990
[46] Yoh-Han Pao. Adaptive Pattern Recognition and Neural Networks. Addison-wesley, 1989
[47] 陈燕庆，鹿洁．神经网络理论及其在控制工程中的应用．西安：西北工业大学，1991
[48] E. Saber, A.M. Tekalp, R. Eschbach, et al. Automatic image annotation using adaptive color classification. Computer Models and Image Processing, 1996, 58(2):115-126
[49] J. Yen, F. Chang, S. Chang. A new criterion for automatic processing, 1995, 4(3):370-378
[50] L. Hayat, M.Fleury, A.F. Clark. Candidate functions for a parallel multi-level thresholding technique. Graphical Models and Image Processing, 1996, 58(4):360-381
[51] Baoxin Li, Rama Chellappa, Oinfen Zheng. Model-Based Temporal Object Verification Using Video. IEEETrans. Image Processing, 2001(6): 897-908
[52] ZiquanHong. Algebraic feature extraction of image for recognition. Pattern Recognition, 1991, 24(3):211-219
[53] 李实，徐旭明，叶榛，等．国际机器人足球比赛及其相关技术．机器人，2000，22(5)：420-426
[54] 高大志．机器人足球——智能机器人的新领域．机器人，1998，20(4)：309—311
[55] 张彦铎，洪炳熔．从信息融合的角度队机器人足球的在认识．机器人技术与应用，2001，1(1)：22-26
[56] 洪炳熔，刘新宇．基于视觉的足球机器人系统．计算机应用研究，2001，1(1)：22-25
[57] 何斌，马天予．Visual C++数字图象处理．人民邮电出版社，2001
[58] 高强，朱安国．Visual C++高级编程技巧．人民邮电出版社，2000
[2] 蔡元龙．模式识别．西安：西北电信工程学院出版社，1986
[3] 朱志刚．数字图像处理．电子工业出版社，1998
[4] 戚飞虎．模式识别与图像处理．上海：上海交通大学出版社，1989
[5] 付京孙．模式识别及其应用．北京：科学出版社，1983
[6] Kenneth R. Castleman. Digital Image Processing. Prentice-Hall, 1996
[7] 胡守仁．神经网络导论．长沙：国防科技大学出版社，1993
[8] 贾云得．机器视觉．北京：科学出版社，2000
[9] Piccioli G, MicheliD, Parodi P, et al. A robust method for road sign detection and recognition. Image and Vision Computing, 1996, 14 (1): 209-233
[10] P. Nordlund, T. Uhlin. Closing the loop: detection and pursuit of a moving observer. Image and Vision Computing, 1996, 14(1):265-275
[11] Clarke J C, Zisserman A. Detection and tracking of independent motion. Image and Vision Computing, 1996, 14(1):565-572
[12] Vieren C, Cabestaing F, Postaire J. Catching moving objects with snakes for motion tracking. Pattern Recognition Letters, 1995, 16(1): 679-685
[13] Hogg D. Model-based vision: A program to see a walking person. Image and Vision Computing, 1995, 1(1):5-20
[14] Rowe S, Blake A. Statistical mosaics for tracking. Image Vision Computing, 1995, 14(1):549-564
[15] Jolly M D. Vehicle Segmentation and Classification Using Deformable Templates. IEEE TRANS. PAMI, 1996, 18(3):225-234.
[16] 杨杨，张田文．基于多分辨率方法的主动轮廓线跟踪算法．计算机学报，1998，21(3)：210-216
[17] 吴一全，朱兆达．图象处理中阈值选取方法30年(1962-1992)的进展(一)．数据采集与处理，1993，8(3)：193-201
[18] 吴一全，朱兆达．图象处理中阈值选取方法30年(1962-1992)的进展(二)．数据采集与处理，1993，8(4)：268-277
[19] 余新平，朱立．一种具有抗噪声干扰的图象边缘提取算法的研究．电子技术应用，1999，25(1)：9-10
[20] 黄伟．人脸识别及其应用研究：[硕士学位论文]．长沙：湖南大学，2001
[21] 彭涤．限制性手写体汉字特征点、笔划段特征的抽取算法研究：[硕士学位论文]．长沙：湖南大学，1993
[22] 伯晓晨，李涛，刘路，等．Matlab工具箱应用指南——信息工程篇．电子工业出版社，2000
[23] 赵景秀，王刚．基于彩色信息区分度的彩色边缘检测．计算机应用，2001，
[24] 姚鸿勋．基于彩色图像的色系坐标变换的面部定位与跟踪法．计算机学报，2002；2(1)：32-35
[25] Casasent D, Linlin Y, Slaski J. Opticalmuti-target subpixel detection and tracking initiation. SPIE, Optical Computing and Nonlinear Materials. 1988, 8(8):12-26
[26] Verri A, Poggio T. Against quantitative optical flow. Proc. of First Conference on Computer Vision. London, 1987
[27] Shahriar Negahdaripout. Revised definition of optical flow: integration of radiometric and geometric cues for dynamic scene analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998, 20(9):961-979
[28] Barron J L, Fleet D J, Beauchemin S S. Performance of optical flow techniques. International Journal of Computer Vision, 1994,12(1):43-47
[29] 张杏，彭晓燕，谭鹤良．足球机器人系统中一种快速模式识别算法的研究与实现．湖南大学学报，2002，6(11)：33-37
[30] Giovanni Adorni, Stefano Cagnoni. Vision-based localization for mobile robots. Robotics and Autonomous Systems, 2001(36):103-119
[31] 佟国峰，赵姝颖，徐心和．提高足球机器人彩色视觉系统实时性的若干图像识别关键技术研究．中南工业大学学报，2000，10(11)：497-500
[32] 欧宗瑛，袁野，张艳珍．基于颜色信息足球机器人视觉跟踪算法．大连理工大学学报，2000，40(6)：729-732
[33] Andrea Bonarini, Paolo Aliverti, Michele Lucioni. An Omni directional Vision Sensor for Fast Tracking for Mobile Robots. IEEE Trans. Instrumentation and Measurement, 2000(49): 509-512
[34] 杨金钟，刘政凯．基于控制点的图象变形方法及其应用．中国图象图形学报，2001，11(6)：1070-1074
[35] Haralick R.M., Shapiro L. G.. Computer and Robot Vision. Mass: Addision-Weslay Publishing Company, 1992
[36] 张勇．机器人足球识别算法与决策仿真：[硕士学位论文]．长沙：湖南大学，2002
[37] R. C. Eberhart, R. W. Dobbins. Neural Network Pe tools. Academic Press,1990
[38] 勒蕃，范俊波．神经网络与神经计算机．西南交通大学，1991
[39] Teuvo kohonen. The Self-organization Map. Proceedings of the IEEE,1990
[40] D.G. Bounds, P.J. Flioyd, B. Matlew, et al. A Multilayer Perceptron Network for the Diagnosis of Low Back Pain. Proc. 2d IEEE Int. Conf. on Neural Networks, London, 1988
[41] P.M. Shea, V. Lin.. Detetical of Explosives in Checked Airline Baggage Using an Artificial Neural Systems. Proc. Int. Joint Conf. on Neural
works, London, 1989
[42] T.J. Sejnowski, C.R. Rosenberg. Parallel Networks that Learn to Pronounce English Text. Complex Systems, 1987
[43] D. Nguyen, B. Widrow. The Truck Backer-upper: An Example of Self-learning in Neural Networks. Proc. Int. Joint Conf. on Neural Networks, Londan, 1989
[44] 庄镇泉，王煦法，王东生．神经网络与神经计算机．北京：科学出版社，1992
[45] Bernard Widrow, Micheal A. Lehr. 30 Years of Adaptive Neural Networks:Perceptron, Madline and Backpropagation. Proc. of the IEEE, 1990
[46] Yoh-Han Pao. Adaptive Pattern Recognition and Neural Networks. Addison-wesley, 1989
[47] 陈燕庆，鹿洁．神经网络理论及其在控制工程中的应用．西安：西北工业大学，1991
[48] E. Saber, A.M. Tekalp, R. Eschbach, et al. Automatic image annotation using adaptive color classification. Computer Models and Image Processing, 1996, 58(2):115-126
[49] J. Yen, F. Chang, S. Chang. A new criterion for automatic processing, 1995, 4(3):370-378
[50] L. Hayat, M.Fleury, A.F. Clark. Candidate functions for a parallel multi-level thresholding technique. Graphical Models and Image Processing, 1996, 58(4):360-381
[51] Baoxin Li, Rama Chellappa, Oinfen Zheng. Model-Based Temporal Object Verification Using Video. IEEETrans. Image Processing, 2001(6): 897-908
[52] ZiquanHong. Algebraic feature extraction of image for recognition. Pattern Recognition, 1991, 24(3):211-219
[53] 李实，徐旭明，叶榛，等．国际机器人足球比赛及其相关技术．机器人，2000，22(5)：420-426
[54] 高大志．机器人足球——智能机器人的新领域．机器人，1998，20(4)：309—311
[55] 张彦铎，洪炳熔．从信息融合的角度队机器人足球的在认识．机器人技术与应用，2001，1(1)：22-26
[56] 洪炳熔，刘新宇．基于视觉的足球机器人系统．计算机应用研究，2001，1(1)：22-25
[57] 何斌，马天予．Visual C++数字图象处理．人民邮电出版社，2001
[58] 高强，朱安国．Visual C++高级编程技巧．人民邮电出版社，2000