智能场景监控系统的算法研究
|
摘要
场景监控系统用于对指定的场景区域进行监视，并把场景内的视频信息传递给监控者，使其能根据相应的情况采取适当措施的系统。监控系统的发展大致经历了三个阶段：模拟场景监控，数字场景监控，智能场景监控系统。智能场景监控系统能把场景内的信息经过智能化的处理后传递给监控者，同时监控者也可通过交互手段对监控过程进行控制，实现对既定场景高效的监控。它融入了视频图像序列理解和计算机视觉的相关知识，在军事、商业、生活上的应用越来越广泛。国际上智能监控领域研究已经达到相当水平，国内则处于刚起步阶段。
本文研究了基本的智能监控系统应具备功能的软件算法，在低层处理算法方面，研究了算法的结构，并针对实际应用中可能出现的问题及算法的时效性等方面进行研究，如对噪声、错误检测的排除、多运动对象的同时处理及合理缩减需处理的信息量等；高层算法方面，基于人工智能的知识和国外该方面的研究，提出了基于规则的事件识别方法；此外，还建立了可扩展的智能监控原型系统，实现了一些简单的检测识别功能，可作为未来的更深入的研究开发的基础；最后本文对原型系统今后如何发展为更为完善和强大的智能场景监控系统提出了若干建议。
Scene supervision and control system is used to supervise a certain scene, transfer the video information in the area to the supervisor, and make supervisor do something according to corresponding condition. There are about three phases during the development of supervision and control system: simulative scene supervision and control system, digital scene supervision and control system, intelligent scene supervision and control system. Intelligent scene supervision and control system can transfer data in the scene that are intelligently processed to the supervisor while the supervisor can in turn control the scene supervision process and confirm an effective supervision and control system in the certain scene. It includes video picture list understand and computer ocular knowledge. It can be widely used in military affairs, commerce, ordinary life. Now research in the intelligent supervisor and control system in the world has reach to an advanced level but in our country it just begins.
The thesis works over the basic software disposal arithmetic that the intelligent supervisor and control system should have. In low level, Study the structure of arithmetic, the problems that may appear in real application and effect of arithmetic such as yawp and error check elimination, dealing synchronously with multi moving body, reducing the
data needed disposal with reason. In high lever, based on knowledge of artificial intelligence and research in this field, bring forward identification way based on rule. What's more, build an extendable antetype of intelligence supervision and control system and realize some simple check and identification function that is the basement of further research. In the end of the thesis, bring forward some suggestions about how to expand the antetype system to a more integrate and power intelligent scene supervision and control system.
本文研究了基本的智能监控系统应具备功能的软件算法，在低层处理算法方面，研究了算法的结构，并针对实际应用中可能出现的问题及算法的时效性等方面进行研究，如对噪声、错误检测的排除、多运动对象的同时处理及合理缩减需处理的信息量等；高层算法方面，基于人工智能的知识和国外该方面的研究，提出了基于规则的事件识别方法；此外，还建立了可扩展的智能监控原型系统，实现了一些简单的检测识别功能，可作为未来的更深入的研究开发的基础；最后本文对原型系统今后如何发展为更为完善和强大的智能场景监控系统提出了若干建议。
Scene supervision and control system is used to supervise a certain scene, transfer the video information in the area to the supervisor, and make supervisor do something according to corresponding condition. There are about three phases during the development of supervision and control system: simulative scene supervision and control system, digital scene supervision and control system, intelligent scene supervision and control system. Intelligent scene supervision and control system can transfer data in the scene that are intelligently processed to the supervisor while the supervisor can in turn control the scene supervision process and confirm an effective supervision and control system in the certain scene. It includes video picture list understand and computer ocular knowledge. It can be widely used in military affairs, commerce, ordinary life. Now research in the intelligent supervisor and control system in the world has reach to an advanced level but in our country it just begins.
The thesis works over the basic software disposal arithmetic that the intelligent supervisor and control system should have. In low level, Study the structure of arithmetic, the problems that may appear in real application and effect of arithmetic such as yawp and error check elimination, dealing synchronously with multi moving body, reducing the
data needed disposal with reason. In high lever, based on knowledge of artificial intelligence and research in this field, bring forward identification way based on rule. What's more, build an extendable antetype of intelligence supervision and control system and realize some simple check and identification function that is the basement of further research. In the end of the thesis, bring forward some suggestions about how to expand the antetype system to a more integrate and power intelligent scene supervision and control system.
引文
1．王润生，图像理解，长沙：国防科技大学出版社，1990
2．闻新，周露，王丹力，神经网络应用设计，北京：科学出版社，1994
3．边肇琪，张学工，模式识别，北京：清华大学出版社，2000
4．何斌，马天予，王运坚，VC++数字图像处理，北京：人民邮电出版社，2001
5．A.MURAT TEKALP，数字视频处理，北京：电子工业出版社，1999
6．王亮，胡卫明，谭铁牛，人运动的视觉分析综述，计算机学报，2002年3月：71—80
7．陈益新，张桂林，实时二值图像去噪声的实现，数据采集与处理，1990年12月：42—46
8．陶毅，龚声蓉，王朝晖，序列图像中运动目标的检测，微电子学与计算机，1993年第11期：34—41
9．白木，周洁，数字视频监控系统，有线电视技术，2002年第18期：51-56
10．张嫣，姚耀文，唐华松，运动图像序列中多目标跟踪的研究与实现，计算机应用研究，2002年：62—69
11．李培华，张田文，主动轮廓线模型(蛇模型)综述，软件学报，2000，11(6)：751-757
12．史立，张兆扬，马然，基于运动跟踪匹配技术的视频对象提取，通信学报，2001年11月，第22卷，第11期：122—129
13．张艳，费耀平，夏利民等，基于活动轮廓模型的胃窦B超序列图像边缘轮廓的提取与跟踪，计算机工程，2000年7月：62—69
14．李智勇，动态图像分析，国防工业出版社，1999：14—18
15．周宏仁，敬忠良，机动目标跟踪，科学出版社，1988：62—67
16．吴琴娣，康华光，吴鸿修，运动图像序列自动解析方法的研究，华中理工大学学报，1994：38—45
17．黄继昌，目标特征提取技术，系统工程与电子技术，1994：29—40
18．孙薇，刘其真，何永保，一种运动物体的动态模糊跟踪新方法，式识别与人工智能，1999：33—37
19．http://www-2.cs.cmu.edu/~vsam/research.html
20. http://www.sinosurveillance.com/
21. Ismail Haritaoglu , David Harwood , Larry S. Davis, Real-Time Surveillance of People and Their Acticities, IEEE Transactions on Pattern Analysis and Machine Intelligence, VOL. 22, NO. 8, AUGUST 2000
22. Mehmet Celenk and Hasnain Reza, Movie Objuect Tracking Using Local Windos, Electrical and Computer Engineering Eepartment, 45701
23. OpenCV Reference Manual, Intel Corporation, 1999-2001
24. An-Christophe Baillie, Jean-Gabriel Ganasca, Action Categorization from Video Sequences
25. Erey Mark Siskind and Q. Morris, A Maximum-likelihood approach to visual event classification, Proceedings of the Fourth European Conference on Computer Vision. Springer-Verlag, 1996.
27. J.M. Siskind, Unsupervised learning of visually observed events, Proceedings of the AAAI Fall Symposium Workshop on Learning Complex Behaviors in Adaptive Intelligent Systems, pp. 82-83, November 1996.
30. Lipton, Local Application of Optic Flow to Analyse Rigid versus Non-Rigid Motion, in ICCV Workshop on Frame-Rate Vision, Corfu, Greece, September 1999
32. Dellaert and Collins, Fast Image-Based Tracking by Selective Pixel Integration, in ICCV Workshop on Frame-Rate Vision, Corfu, Greece, September 1999
33. Wixson and Hansen , Detecting Salient Motion by Accumulating Directionally-Consistent Flow, ICCV, 1999, pp. 797-804
34. Collins and Tsin, Calibration of an Outdoor Active Camera System, IEEE Computer Vision and Pattern Recognition (CVPR99), Fort Collins, CO, June 23-25, 1999, pp. 528-534
35. Collins, Lipton and Kanade, A System for Video Surveillance and Monitoring, Proc. American Nuclear Society (ANS) Eighth International Topical Meeting on Robotics and Remote Systems, Pittsburgh, PA, April
25-29, 1999
36. Lipton, Fujiyoshi and Patil, Moving Target Classification and Tracking from Real-time Video, IEEE Workshop on Applications of Computer Vision (WACV), Princeton NJ, October 1998, pp. 8-14.
37. Fujiyoshi and Lipton, Real-time Human Motion Analysis by Image Skeletonization, IEEE Workshop on Applications of Computer Vision (WACV), Princeton NJ, October 1998, pp. 15-21
38. Kanade, Collins, Lipton, Burt, Wixson, Advances in Cooperative Muir\i-Sensor Video Surveillance, DARPA Image Understanding Workshop (IUW), Monterey, CA, November 1998, pp. 3-24
39. Wixson, Eledath, Hansen, Image Alignment for Precise Camera Fixation and Aim IEEE Computer Vision and Pattern Recognition (CVPR), 1998, pp. 708-715
40. Collins, Tsin, Miller and Lipton, Using a DEM to Determine Geospatial Object Trajectories, DARPA Image Understanding Workshop (IUW), Monterey, CA, November 1998, pp. 115-122..
41. Selinger and Wilson, Classifying Moving Objects as Rigid or Non-Rigid without Correspondences DARPA Image Understanding Workshop (IUW), Monterey, CA, November 1998, pp. 341-348
42. Kanade, Collins, Lipton, Anandan, Burt, Wixson, Cooperative Multi-Sensor Video Surveillance, DARPA Image Understanding Workshop (IUW), New Orleans, LA, May 1997, pp. 3-10.
2．闻新，周露，王丹力，神经网络应用设计，北京：科学出版社，1994
3．边肇琪，张学工，模式识别，北京：清华大学出版社，2000
4．何斌，马天予，王运坚，VC++数字图像处理，北京：人民邮电出版社，2001
5．A.MURAT TEKALP，数字视频处理，北京：电子工业出版社，1999
6．王亮，胡卫明，谭铁牛，人运动的视觉分析综述，计算机学报，2002年3月：71—80
7．陈益新，张桂林，实时二值图像去噪声的实现，数据采集与处理，1990年12月：42—46
8．陶毅，龚声蓉，王朝晖，序列图像中运动目标的检测，微电子学与计算机，1993年第11期：34—41
9．白木，周洁，数字视频监控系统，有线电视技术，2002年第18期：51-56
10．张嫣，姚耀文，唐华松，运动图像序列中多目标跟踪的研究与实现，计算机应用研究，2002年：62—69
11．李培华，张田文，主动轮廓线模型(蛇模型)综述，软件学报，2000，11(6)：751-757
12．史立，张兆扬，马然，基于运动跟踪匹配技术的视频对象提取，通信学报，2001年11月，第22卷，第11期：122—129
13．张艳，费耀平，夏利民等，基于活动轮廓模型的胃窦B超序列图像边缘轮廓的提取与跟踪，计算机工程，2000年7月：62—69
14．李智勇，动态图像分析，国防工业出版社，1999：14—18
15．周宏仁，敬忠良，机动目标跟踪，科学出版社，1988：62—67
16．吴琴娣，康华光，吴鸿修，运动图像序列自动解析方法的研究，华中理工大学学报，1994：38—45
17．黄继昌，目标特征提取技术，系统工程与电子技术，1994：29—40
18．孙薇，刘其真，何永保，一种运动物体的动态模糊跟踪新方法，式识别与人工智能，1999：33—37
19．http://www-2.cs.cmu.edu/~vsam/research.html
20. http://www.sinosurveillance.com/
21. Ismail Haritaoglu , David Harwood , Larry S. Davis, Real-Time Surveillance of People and Their Acticities, IEEE Transactions on Pattern Analysis and Machine Intelligence, VOL. 22, NO. 8, AUGUST 2000
22. Mehmet Celenk and Hasnain Reza, Movie Objuect Tracking Using Local Windos, Electrical and Computer Engineering Eepartment, 45701
23. OpenCV Reference Manual, Intel Corporation, 1999-2001
24. An-Christophe Baillie, Jean-Gabriel Ganasca, Action Categorization from Video Sequences
25. Erey Mark Siskind and Q. Morris, A Maximum-likelihood approach to visual event classification, Proceedings of the Fourth European Conference on Computer Vision. Springer-Verlag, 1996.
27. J.M. Siskind, Unsupervised learning of visually observed events, Proceedings of the AAAI Fall Symposium Workshop on Learning Complex Behaviors in Adaptive Intelligent Systems, pp. 82-83, November 1996.
30. Lipton, Local Application of Optic Flow to Analyse Rigid versus Non-Rigid Motion, in ICCV Workshop on Frame-Rate Vision, Corfu, Greece, September 1999
32. Dellaert and Collins, Fast Image-Based Tracking by Selective Pixel Integration, in ICCV Workshop on Frame-Rate Vision, Corfu, Greece, September 1999
33. Wixson and Hansen , Detecting Salient Motion by Accumulating Directionally-Consistent Flow, ICCV, 1999, pp. 797-804
34. Collins and Tsin, Calibration of an Outdoor Active Camera System, IEEE Computer Vision and Pattern Recognition (CVPR99), Fort Collins, CO, June 23-25, 1999, pp. 528-534
35. Collins, Lipton and Kanade, A System for Video Surveillance and Monitoring, Proc. American Nuclear Society (ANS) Eighth International Topical Meeting on Robotics and Remote Systems, Pittsburgh, PA, April
25-29, 1999
36. Lipton, Fujiyoshi and Patil, Moving Target Classification and Tracking from Real-time Video, IEEE Workshop on Applications of Computer Vision (WACV), Princeton NJ, October 1998, pp. 8-14.
37. Fujiyoshi and Lipton, Real-time Human Motion Analysis by Image Skeletonization, IEEE Workshop on Applications of Computer Vision (WACV), Princeton NJ, October 1998, pp. 15-21
38. Kanade, Collins, Lipton, Burt, Wixson, Advances in Cooperative Muir\i-Sensor Video Surveillance, DARPA Image Understanding Workshop (IUW), Monterey, CA, November 1998, pp. 3-24
39. Wixson, Eledath, Hansen, Image Alignment for Precise Camera Fixation and Aim IEEE Computer Vision and Pattern Recognition (CVPR), 1998, pp. 708-715
40. Collins, Tsin, Miller and Lipton, Using a DEM to Determine Geospatial Object Trajectories, DARPA Image Understanding Workshop (IUW), Monterey, CA, November 1998, pp. 115-122..
41. Selinger and Wilson, Classifying Moving Objects as Rigid or Non-Rigid without Correspondences DARPA Image Understanding Workshop (IUW), Monterey, CA, November 1998, pp. 341-348
42. Kanade, Collins, Lipton, Anandan, Burt, Wixson, Cooperative Multi-Sensor Video Surveillance, DARPA Image Understanding Workshop (IUW), New Orleans, LA, May 1997, pp. 3-10.