电视制导目标跟踪算法的研究
|
摘要
基于可见光成像的自动目标跟踪技术是一种重要的计算机图像处理应用技术,目前被广泛应用于武器制导、交通监测、机器人视觉、以及导航等领域。自动目标跟踪技术实现的关键是综合考虑算法的稳定性、有效性及实时性。相对于简单背景,地面复杂背景下的目标更难于提取其形状等特征,因此在选择跟踪算法上主要研究了基于图像匹配算法的跟踪方法,其理论基础是相似度。
文中首先介绍了电视制导技术的概况及成像跟踪技术的发展现状,然后讨论了图像匹配的各种算法,接着对图像匹配的各种加速算法进行了研究,并对传统的分层搜索算法进行了改进。针对模板匹配计算量大的问题,采用了一种先用轨迹预测法预测目标位置,再用改进的加速相关匹配算法的方法,有效地提高了运算速度。详细分析了目标跟踪算法的一些参数确定。针对相关匹配算法跟踪误差的积累会导致模板漂移的问题,研究了基于模板质心的模板检验策略和动态加权模板更新策略。仿真试验结果表明,本文提出的自动目标跟踪算法具有较好的匹配精度和稳定性。
详细说明了基于DSP的电视制导目标跟踪仿真系统的硬件平台和跟踪软件设计,并给出了软件流程图。完成了系统的图像处理和图像显示两个重要模块的软件设计和调试工作,实验证明该自动目标跟踪系统的性能可靠,跟踪效果较好。
Automatic object tracking technology based on visible light imaging is oneof the most important computer image processing technology applications. It hasbeen extensively applied in various fields such as weapon guiding, trafficdetection, robot vision, navigation and so on. The key points for Automatic objecttracking technology are to consider stabilization, efficiency, real time of thearithmetic at the same time. By contrast with the simple background, it is hard topick-up object’s various characteristics for example its shape from the complexground background. So the tracking methods based on image matching arithmeticare studied mostly. Its academic base is comparability measurement.
In this paper, general situation of TV-guided and development actuality ofimage tracking technology are introduced. Then some algorithms of imagematching are discussed. Different accelerating algorithms of image matching arestudied and improved traditional searching algorithm is studied. The modelmatching technique is confronted with large computational quantity, so a methodis adopted, in which the location of the target is forecasted based on trackprediction first, then a improved accelerated correlation matching arithmetic isapplied. By using this method, the matching time can be greatly reduced. Theauthor analyses some parameters of object tracking arithmetic detailedly. For thecumulated errors of correlation arithmetic leads to model flying, dynamic modelupdating strategy and modifying strategy based on model’s mass center arestudied. Results of simulation experiment show that the tracking arithmetic thispaper gives has the better matching precision and stability.
The hardware platform and software design of TV-guided object trackingsimulation system based on DSP are described. The software flow of the objecttracking system is given. The software design of image processing module andimage display module and system software debug is completed. Experimentalresults show the automatic object tracking system works reliable, effective.
文中首先介绍了电视制导技术的概况及成像跟踪技术的发展现状,然后讨论了图像匹配的各种算法,接着对图像匹配的各种加速算法进行了研究,并对传统的分层搜索算法进行了改进。针对模板匹配计算量大的问题,采用了一种先用轨迹预测法预测目标位置,再用改进的加速相关匹配算法的方法,有效地提高了运算速度。详细分析了目标跟踪算法的一些参数确定。针对相关匹配算法跟踪误差的积累会导致模板漂移的问题,研究了基于模板质心的模板检验策略和动态加权模板更新策略。仿真试验结果表明,本文提出的自动目标跟踪算法具有较好的匹配精度和稳定性。
详细说明了基于DSP的电视制导目标跟踪仿真系统的硬件平台和跟踪软件设计,并给出了软件流程图。完成了系统的图像处理和图像显示两个重要模块的软件设计和调试工作,实验证明该自动目标跟踪系统的性能可靠,跟踪效果较好。
Automatic object tracking technology based on visible light imaging is oneof the most important computer image processing technology applications. It hasbeen extensively applied in various fields such as weapon guiding, trafficdetection, robot vision, navigation and so on. The key points for Automatic objecttracking technology are to consider stabilization, efficiency, real time of thearithmetic at the same time. By contrast with the simple background, it is hard topick-up object’s various characteristics for example its shape from the complexground background. So the tracking methods based on image matching arithmeticare studied mostly. Its academic base is comparability measurement.
In this paper, general situation of TV-guided and development actuality ofimage tracking technology are introduced. Then some algorithms of imagematching are discussed. Different accelerating algorithms of image matching arestudied and improved traditional searching algorithm is studied. The modelmatching technique is confronted with large computational quantity, so a methodis adopted, in which the location of the target is forecasted based on trackprediction first, then a improved accelerated correlation matching arithmetic isapplied. By using this method, the matching time can be greatly reduced. Theauthor analyses some parameters of object tracking arithmetic detailedly. For thecumulated errors of correlation arithmetic leads to model flying, dynamic modelupdating strategy and modifying strategy based on model’s mass center arestudied. Results of simulation experiment show that the tracking arithmetic thispaper gives has the better matching precision and stability.
The hardware platform and software design of TV-guided object trackingsimulation system based on DSP are described. The software flow of the objecttracking system is given. The software design of image processing module andimage display module and system software debug is completed. Experimentalresults show the automatic object tracking system works reliable, effective.
引文
1范金荣.制导炸弹发展综述.现代防御技术. 2004, 32(3):28~31
2黄世奇,禹春来,刘代志,钱昌松.成像精确制导技术分析与研究.导弹与航天运载技术. 2005, (5):20~21
3艾金蔚,刘克.视频序列中运动目标跟踪新方法.北京科技大学学报,2006, (28):195~198
4 Nguyen H T, Worring M, Rvan den boomgaard. Occlusion Robust AdaptiveTemplate Tracking. Proc IEEE Conf on Computer Vision. 2001, (1):23~25
5刘刚,苏秀琴,胡晓东,郝伟,杨小军.高速电视下基于自适应阈值的实时图像跟踪.光子学报. 2005, 34(8):1262~1265
6 M. Arulampalam, S. Maskell, N. Gordon, T. Clapp. A Tutorial on ParticleFilters for Online Nonlinear/Non-Gaussian Bayesian Tracking. IEEETransactions on Signal Processing. 2002, 50(2):174~189
7 Violap, Jonesm. Rapid object detection using a boosted cascade of simplefeature. IEEE Computer Society Conference on Computer Vision and PatternRecognition. 2001, 1(12):512~518
8刘畅,向健勇,杨斌利.一种新的基于梯度特征的快速相关匹配算法.激光与红外. 2004, 34(4):305~307
9 Nickels, K., and Hutchinson, S. Estimating Uncertainty in SSD-based FeatureTracking. Image Vis. Computer. 2002, 20(1):47~58
10邵文坤,黄爱民,韦庆.动态场景下的运动目标跟踪方法研究.计算机仿真. 2006, 23(5):181~184
11 Glodberg D E. Genetic Algorithms in Search Optimization and MatchingLearning. Addsion Wesley Publishing Company, 1989:16~23
12 Chen, M.J., et al. A new Block-matching Criterion for Motion Estimation andits Implementation. IEEE Trans. Circuits Syst Video Technol. 1995, 5(3):231~236
13 Natarajan, B., Bhaskaran, V., and Konstantinides, K. Low-complexity BlockbasedMotion Estimation via One-bit Transforms. IEEE Trans. Circuits Syst.Video Technol. 1997, 7(4):702~706
14黄旭华.基于灰度的图像快速匹配算法.战术导弹控制技. 2005, (4):25~27
15罗诗途,张玘,罗飞路,王艳玲.基于自适应修正的相关跟踪算法.仪器仪表学报. 2004, (8):633~635
16杨宜禾,周维真.成像跟踪技术导论.西安电子科技大学出社,1992:25~30
17 Aiming Huang, Zheng Gao, Bing Dai and Li Luo. Real-time Object Matching.SPIE Vol. 3460:755~760
18 X. Y. Ren and Y. T. Liao. A New Correlation Tracking Method. Journal ofImage and Graphics. 2002, 7(5):553~557
19 Baker E S, De Groat R D. A Correlation-based Subspace Tracking Algorithm.Signal Processing, IEEE Transaction on. 1998, 46(11):3112~3116
20 Eklund M V, Ravichandran G, Trived M M. Real-time Visual TrackingUsing Correlation Techniques. Applications of Computer Vision Proceedingof the Second IEEE Workshop on. 1994:256 ~263.
21 Peter Aschwanden. Method for Real-time Tracking of Uncooperative Targets.Proc.Of the int . Conf On ASSP. 1989, (3 ):1560 ~1563
22周丰乐,徐向民,肖跃,周娇.一种新的二值图像目标轮廓跟踪算法.微计算机信息. 2007, 23(2):259 ~261
23 Patrick Brigger, Jeff Hoeg, Michael Unser. B-spline Snakes: a Flexible Toolfor Parametric Contour Detection. IEEE Transactions on Image Processing.2000, 9(9):1484~1496
24李弼程,彭天强,彭波.智能图像处理技术.电子工业出版社, 2004:245~246 239~244
25 H. J. Yang. Design and Realization of a New Correlation TrackingAlgorithm. Journal of Infrared and Millimeter Waves. 2000, 19(5):377~380
26陈建华,张宇河.基于边缘的自适应相关跟踪算法.视频技术应用与工程.2004, (4):67~68
27江泽涛,赵荣椿,黎明.基于相关的自适应阈值图像序列跟踪方法.西北工业大学学报. 2005, 23(6):804~807
28倪军,周进,袁家虎,吴钦章.空间目标序列图像相关跟踪算法.光电工程. 2006, 33(4):1~4
29杨勇智,文远保.一种改进的快速相关跟踪算法.计算机应用. 2005,25(12):2843~2844
30江河平,王平,沈振康.任意指定地面复杂场景图像目标的去均值相关跟踪算法.信号处理. 2006, 22(1):24~27
31徐瑞鑫.复杂背景下实时目标跟踪与识别研究.中国科学院硕士论文.2002:21~23
32沈海滨,赖汝.基于图像中心矩的快速模板匹配算法.计算机应用. 2004,24(11):113~115
33雍杨,王敬儒,张启衡.基于塔形的快速相关跟踪算法.光电工程. 2003,30(6):11~12
34 Cavicchio D J. Adaptive search using simulated evolution. DoctoralDissertation University of Michigan Ann Arbor.1970
35 Cavicchio D J. Reproductive Adaptive Plants Proceedings of the Acm. 1972Annual Conference.1972:1~11
36吉书鹏.一种小波多分辨率分析的多算子动态场景目标跟踪算法研究.光学技术. 2004, 30(6):646~652
37张旭光,王明佳,王延杰.自适应选取阈值的三层快速匹配算法.光电工程. 2005, 32(6):83~86
38胡硕,朱明,宋华军,李建军.一种新的基于TMS320C6416平台下的相关跟踪算法.电子器件. 2006, 29(2):416~419
39崔青松.目标检测跟踪技术研究及DSP实现.中北大学硕士论文. 2006:20~28
40季昱,林俊超,于本喜. DSP嵌入式应用系统开发典型实例.中国电力出版社, 2005:53~56 69~70
41陆宗骐. C/C++图像处理编程.清华大学出版社, 2005:38~47 216~219
42王家文,李仰军. MATLAB7.0图形图像处理.国防工业出版社, 2006:44~47 122~127 253~257
43 Rafael C.Gonzalez, Richard E. Woods.数字图像处理.阮秋奇,阮宇智.第二版.电子工业出版社, 2003:462~474
44 Gharavi, H.and Mills, M. Block Matching Motion Estimation Algorithm newResults. IEEE Trans. Circuits Systems. 1990, 37(5):649~651
45 Lin Cheng Wang, Sandor Z Der, Nasser M Nasrabadi. Automatic TargetRecognition Using a Feature-Decomposition and Data-DecompositionModular Neural Network. IEEE Transaction on Image Processing. 1998, 7(8)
46 Andrew Blake. Active Vision. Cambridge: The MIT Press, 1992:39~58
47 Liu Weiyi, Wang Zhaoqi, Mu Guoguangetal. Three Dimension Reconstructionby Space Image Fusion. Acta Optic Sinic. 2000, 20(3): 363~369
48周晓亮.基于DSP的数字视频图像获取与处理技术研究.华中科技大学硕士论文. 2004:8~12
2黄世奇,禹春来,刘代志,钱昌松.成像精确制导技术分析与研究.导弹与航天运载技术. 2005, (5):20~21
3艾金蔚,刘克.视频序列中运动目标跟踪新方法.北京科技大学学报,2006, (28):195~198
4 Nguyen H T, Worring M, Rvan den boomgaard. Occlusion Robust AdaptiveTemplate Tracking. Proc IEEE Conf on Computer Vision. 2001, (1):23~25
5刘刚,苏秀琴,胡晓东,郝伟,杨小军.高速电视下基于自适应阈值的实时图像跟踪.光子学报. 2005, 34(8):1262~1265
6 M. Arulampalam, S. Maskell, N. Gordon, T. Clapp. A Tutorial on ParticleFilters for Online Nonlinear/Non-Gaussian Bayesian Tracking. IEEETransactions on Signal Processing. 2002, 50(2):174~189
7 Violap, Jonesm. Rapid object detection using a boosted cascade of simplefeature. IEEE Computer Society Conference on Computer Vision and PatternRecognition. 2001, 1(12):512~518
8刘畅,向健勇,杨斌利.一种新的基于梯度特征的快速相关匹配算法.激光与红外. 2004, 34(4):305~307
9 Nickels, K., and Hutchinson, S. Estimating Uncertainty in SSD-based FeatureTracking. Image Vis. Computer. 2002, 20(1):47~58
10邵文坤,黄爱民,韦庆.动态场景下的运动目标跟踪方法研究.计算机仿真. 2006, 23(5):181~184
11 Glodberg D E. Genetic Algorithms in Search Optimization and MatchingLearning. Addsion Wesley Publishing Company, 1989:16~23
12 Chen, M.J., et al. A new Block-matching Criterion for Motion Estimation andits Implementation. IEEE Trans. Circuits Syst Video Technol. 1995, 5(3):231~236
13 Natarajan, B., Bhaskaran, V., and Konstantinides, K. Low-complexity BlockbasedMotion Estimation via One-bit Transforms. IEEE Trans. Circuits Syst.Video Technol. 1997, 7(4):702~706
14黄旭华.基于灰度的图像快速匹配算法.战术导弹控制技. 2005, (4):25~27
15罗诗途,张玘,罗飞路,王艳玲.基于自适应修正的相关跟踪算法.仪器仪表学报. 2004, (8):633~635
16杨宜禾,周维真.成像跟踪技术导论.西安电子科技大学出社,1992:25~30
17 Aiming Huang, Zheng Gao, Bing Dai and Li Luo. Real-time Object Matching.SPIE Vol. 3460:755~760
18 X. Y. Ren and Y. T. Liao. A New Correlation Tracking Method. Journal ofImage and Graphics. 2002, 7(5):553~557
19 Baker E S, De Groat R D. A Correlation-based Subspace Tracking Algorithm.Signal Processing, IEEE Transaction on. 1998, 46(11):3112~3116
20 Eklund M V, Ravichandran G, Trived M M. Real-time Visual TrackingUsing Correlation Techniques. Applications of Computer Vision Proceedingof the Second IEEE Workshop on. 1994:256 ~263.
21 Peter Aschwanden. Method for Real-time Tracking of Uncooperative Targets.Proc.Of the int . Conf On ASSP. 1989, (3 ):1560 ~1563
22周丰乐,徐向民,肖跃,周娇.一种新的二值图像目标轮廓跟踪算法.微计算机信息. 2007, 23(2):259 ~261
23 Patrick Brigger, Jeff Hoeg, Michael Unser. B-spline Snakes: a Flexible Toolfor Parametric Contour Detection. IEEE Transactions on Image Processing.2000, 9(9):1484~1496
24李弼程,彭天强,彭波.智能图像处理技术.电子工业出版社, 2004:245~246 239~244
25 H. J. Yang. Design and Realization of a New Correlation TrackingAlgorithm. Journal of Infrared and Millimeter Waves. 2000, 19(5):377~380
26陈建华,张宇河.基于边缘的自适应相关跟踪算法.视频技术应用与工程.2004, (4):67~68
27江泽涛,赵荣椿,黎明.基于相关的自适应阈值图像序列跟踪方法.西北工业大学学报. 2005, 23(6):804~807
28倪军,周进,袁家虎,吴钦章.空间目标序列图像相关跟踪算法.光电工程. 2006, 33(4):1~4
29杨勇智,文远保.一种改进的快速相关跟踪算法.计算机应用. 2005,25(12):2843~2844
30江河平,王平,沈振康.任意指定地面复杂场景图像目标的去均值相关跟踪算法.信号处理. 2006, 22(1):24~27
31徐瑞鑫.复杂背景下实时目标跟踪与识别研究.中国科学院硕士论文.2002:21~23
32沈海滨,赖汝.基于图像中心矩的快速模板匹配算法.计算机应用. 2004,24(11):113~115
33雍杨,王敬儒,张启衡.基于塔形的快速相关跟踪算法.光电工程. 2003,30(6):11~12
34 Cavicchio D J. Adaptive search using simulated evolution. DoctoralDissertation University of Michigan Ann Arbor.1970
35 Cavicchio D J. Reproductive Adaptive Plants Proceedings of the Acm. 1972Annual Conference.1972:1~11
36吉书鹏.一种小波多分辨率分析的多算子动态场景目标跟踪算法研究.光学技术. 2004, 30(6):646~652
37张旭光,王明佳,王延杰.自适应选取阈值的三层快速匹配算法.光电工程. 2005, 32(6):83~86
38胡硕,朱明,宋华军,李建军.一种新的基于TMS320C6416平台下的相关跟踪算法.电子器件. 2006, 29(2):416~419
39崔青松.目标检测跟踪技术研究及DSP实现.中北大学硕士论文. 2006:20~28
40季昱,林俊超,于本喜. DSP嵌入式应用系统开发典型实例.中国电力出版社, 2005:53~56 69~70
41陆宗骐. C/C++图像处理编程.清华大学出版社, 2005:38~47 216~219
42王家文,李仰军. MATLAB7.0图形图像处理.国防工业出版社, 2006:44~47 122~127 253~257
43 Rafael C.Gonzalez, Richard E. Woods.数字图像处理.阮秋奇,阮宇智.第二版.电子工业出版社, 2003:462~474
44 Gharavi, H.and Mills, M. Block Matching Motion Estimation Algorithm newResults. IEEE Trans. Circuits Systems. 1990, 37(5):649~651
45 Lin Cheng Wang, Sandor Z Der, Nasser M Nasrabadi. Automatic TargetRecognition Using a Feature-Decomposition and Data-DecompositionModular Neural Network. IEEE Transaction on Image Processing. 1998, 7(8)
46 Andrew Blake. Active Vision. Cambridge: The MIT Press, 1992:39~58
47 Liu Weiyi, Wang Zhaoqi, Mu Guoguangetal. Three Dimension Reconstructionby Space Image Fusion. Acta Optic Sinic. 2000, 20(3): 363~369
48周晓亮.基于DSP的数字视频图像获取与处理技术研究.华中科技大学硕士论文. 2004:8~12