EMCV库扩充及车辆检测系统的嵌入式实现
摘要
随着科技的进步,车辆给人们的生活带来便利的同时也带来了社会问题,交通安全成为人们普遍关心的话题,同时对于交通领域的相关技术的要求也在不断的提高。车辆检测技术是车辆跟踪、车型分类、车速检测等智能交通方面研究的前提和基础,也是前方碰撞警示技术以及主动安全技术等至关重要的一步。车辆检测技术是指从复杂的交通信息中将目标车辆检测出来,即目标提取的过程。
本篇论文分析比较常用的目标提取算法,将模式识别中的Adboost算法运用到车辆检测系统中,完成了可以不用通过背景建模、更新来实现车辆检测的功能。Adboost算法分为Discrete Adboost、Real Adboost、LogitBoost以及GentleBoost四种,每一种算法都有各自的优点和不足,本文选择了GentleBoost算法,并且证明了该算法在车辆检测系统中的应用具有可行性和准确性。
本文对EMCV库进行了扩充,实现了基于GentleBoost的车辆检测算法。本文实现的车辆检测系统是基于DSP系统平台的,通过结合系统的硬件部分设计了系统软件部分,包括存储器分配、DSP/BIOS系统框架构建、软件驱动以及系统线程间的通信等问题。针对实际应用情况,建立了车辆检测分类器,经过大量的实验和测试,确定了选取正负样本、分类器训练过程中的原则和注意事项,以及参数如何设置等问题；使EMCV库函数可以在DSP上运行,将前面设计的适合该系统的分类器加载到该系统中,完成车辆检测系统的视频口采集和显示设置,建立视频采集和视频图像处理显示任务,采集视频流将车辆信息在视频帧图像上标记出来,使车辆检测算法在DSP系统平台上实现。
With the progress of technology, vehicle not only brings convenience to human, but also brings social problems. Traffic security become a widespread concern topic, meanwhile the demand of transportation related technologies is increasingly improved. Vehicle detection is the premise and foundation of the intelligent transportation research, such as vehicle tracking, vehicle models classification, vehicle speed test and so on, which is also a vital step of the forward collision warming technology and active safety technology. Vehicle detection is detecting the vehicle in complicated conditions.
In the thesis, the general moving object extracted algorithms are compared, and Adboost algorithm that belong to pattern recognition is applied to the vehicle detection system, which can achieve vehicle detection without objective extracting, background modeling or updating. Adboost algorithm is divided into Discrete Adboost algorithm, Real Adboost, LogitBoost and GentleBoost, each algorithm has its own advantages and disadvantages. Finally, GentleBoost algorithm is selected to implement the vehicle system, and its accuracy and feasibility are approved in the thesis.
In the thesis, EMCV library is expanded, and the vehicle system is achieved based on GentleBoost algorithm. The vehicle detection system is based on DSP system, according to the hardware module. And the software module is designed, which include memory allocation, DSP/BIOS system framework building, software driver and communications between the system threads. Considered the actual application, the vehicle detection classifier is established, after a great deal of experiments and tests, the author identify the principles and consideration of positive and negative samples selecting and the classifier training processing as well as how to set the classifier parameters, etc, and test Adboost algorithm performance on OpenCV platform.,the author implement the vehicle detection system by running the EMCV library on DSP, loading the former classifier, setting the parameters of input/output video port, creating the task of video capturing, displaying and video image processing, capturing the video stream and marking the vehicle information.
本篇论文分析比较常用的目标提取算法,将模式识别中的Adboost算法运用到车辆检测系统中,完成了可以不用通过背景建模、更新来实现车辆检测的功能。Adboost算法分为Discrete Adboost、Real Adboost、LogitBoost以及GentleBoost四种,每一种算法都有各自的优点和不足,本文选择了GentleBoost算法,并且证明了该算法在车辆检测系统中的应用具有可行性和准确性。
本文对EMCV库进行了扩充,实现了基于GentleBoost的车辆检测算法。本文实现的车辆检测系统是基于DSP系统平台的,通过结合系统的硬件部分设计了系统软件部分,包括存储器分配、DSP/BIOS系统框架构建、软件驱动以及系统线程间的通信等问题。针对实际应用情况,建立了车辆检测分类器,经过大量的实验和测试,确定了选取正负样本、分类器训练过程中的原则和注意事项,以及参数如何设置等问题；使EMCV库函数可以在DSP上运行,将前面设计的适合该系统的分类器加载到该系统中,完成车辆检测系统的视频口采集和显示设置,建立视频采集和视频图像处理显示任务,采集视频流将车辆信息在视频帧图像上标记出来,使车辆检测算法在DSP系统平台上实现。
With the progress of technology, vehicle not only brings convenience to human, but also brings social problems. Traffic security become a widespread concern topic, meanwhile the demand of transportation related technologies is increasingly improved. Vehicle detection is the premise and foundation of the intelligent transportation research, such as vehicle tracking, vehicle models classification, vehicle speed test and so on, which is also a vital step of the forward collision warming technology and active safety technology. Vehicle detection is detecting the vehicle in complicated conditions.
In the thesis, the general moving object extracted algorithms are compared, and Adboost algorithm that belong to pattern recognition is applied to the vehicle detection system, which can achieve vehicle detection without objective extracting, background modeling or updating. Adboost algorithm is divided into Discrete Adboost algorithm, Real Adboost, LogitBoost and GentleBoost, each algorithm has its own advantages and disadvantages. Finally, GentleBoost algorithm is selected to implement the vehicle system, and its accuracy and feasibility are approved in the thesis.
In the thesis, EMCV library is expanded, and the vehicle system is achieved based on GentleBoost algorithm. The vehicle detection system is based on DSP system, according to the hardware module. And the software module is designed, which include memory allocation, DSP/BIOS system framework building, software driver and communications between the system threads. Considered the actual application, the vehicle detection classifier is established, after a great deal of experiments and tests, the author identify the principles and consideration of positive and negative samples selecting and the classifier training processing as well as how to set the classifier parameters, etc, and test Adboost algorithm performance on OpenCV platform.,the author implement the vehicle detection system by running the EMCV library on DSP, loading the former classifier, setting the parameters of input/output video port, creating the task of video capturing, displaying and video image processing, capturing the video stream and marking the vehicle information.
引文
[1]http://auto.qq.com/a/20110318/000021.html
[2]林学言,王宏等.计算机视觉：一种现代的方法.北京：电子工业出版社,2004.
[3]刘瑞祯,于仕琪. OpenCV教程——基础篇.北京：北京航空航天人学出版社,2007.
[4]http://www.opencv.org.cn/forum/viewforum.php?f=9
[5]吴冬梅,张玉杰.DSP技术及应用.北京：北京大学出版社,2006.
[6]http://www.zhinengjiaotong.com/tech/show-9. html
[7]Leibe B, Leonardis A, Schiele B. Combined object categorization and segmentation with an implicit shape model. European Conference on Computer Vision, Switzerland, 2004:3-10.
[8]Felzenszwalb P, McAllester D, Ramanan D. A Discriminatively Trained, Multiscale, Deformable Part Mode. CVPR, USA,2008:1-8.
[9]Huang Y, Huang K, et al. Enhanced Biologically Inspired Model. CVPR, USA,2008:1-8.
[10]Papageorgiou C, Poggio T. A Trainable System for Object Detection. International Journal of Computer Vision.2000,38(1):15-33.
[11]Kapoor A, Winn J. Located Hidden Random Fields:Learning Discriminative Parts for Object Detection. European Conference on Computer Vision, Austria,2006:2-13
[12]Furst L, Fidler S, Leonardis A. Selecting Features for Object Detection Using an Adboost compatible Evaluation Function.2008, Pattern Recognition Letters(29): 1603-1612.
[13]J. Fang, G. Qiu. Car/non-car classification in an informative sample subspace. In Proceedings of ICPR, Hong kong,2006:962-965.
[14]Z. Zhu, Y. Zhao, H. Lu. Sequential architecture for efficient car detection. IEEE Conference on Computer Vision and Pattern Recognition, USA,2007:1-8.
[15]http://www.docin.com/p-5289933.html
[16]Haritaoglu I, Harwood D, Davis L. S. Real-time surveillance of people and the activities. IEEE Trans Pattern Anal.2000, Machine Intel(22):809-830.
[17]C. Wren, A. Azarbaygaui. Pfinder:Real-time tracking of the human body. IEEE Tran Pattern Anal.1997, Machine Intel(19):780-785.
[18]A. J. SCHOFIELD, A. MEHTA, T. J. STOXHAM. A system for counting people in video images using neural networks to identify the background scene. Pattern Recognition,1996, 29(8):1421-1428.
[19]谷口庆治.数字图像处理------应用篇.北京：科学出版社,2002.
[20]鲁一杰,赵攀,彭强等.基于DSP的视频车辆实时检测系统的实现.计算机应用研究.2009,26(2)：758-759.
[21]Rainer Lien hart, Jochen Maydt. An Extended Set of Haar-like Features for Rapid Object Detection. IEEE ICIP, New York,2002:900-903.
[22]Paul Viola, Michael Jones. Robust Real-time Object Detection. IEEE International Conference on Computer Vision, Canada,2001:1-22.
[23]唐立群,郭庆昌,李永华.数字图像模式识别方法分析.哈尔滨：哈尔滨工程大学出版社,2008.
[24]Freund, Schapire. A decision-theoretic generalization of on-line learning and an application to boosting. In Machine Learning:Proceedings of the Thirteenth International Conference, SanFrancisco,1996:148-156.
[25]Robert E, Yoram Singer. Improved Boosting Algorithms Using Confidence-rated Predictions. Machine Learning, Netherlands,1999:297-336.
[26]Jerome Friedman, Trevor Hastie, Robert Tibshirani. Additive Logistic Regression: a Statistical View of Boosting. Annals of Statistics,2000,28(2):337-407.
[27]Lior W, Lan M. Robust boosting for learning from few example [J]. IEEE Conference on Computer Vision and Pattern Recognition, San Diego,2005:395-364.
[28]Chuck White等著.周生炳等译.XML从入门到精通(黄金版).北京：电子工业出版社,2002.
[29]王跃宗,刘京会.TMS320DM642DSP应用系统设计与开发.北京：人民邮电出版社,2009.
[30]李芳慧,王飞,何佩琨.TM320C6000系列DSP原理与应用(第2版)[M].北京：电子工业出版社,2003.
[31]隋忠磊.基于DM642的数字图像采集处理系统[D].内蒙古：内蒙古工业大学,2007.
[32]李保民.基于DM642的视频图像实时处理模块的设计与实现[D].武汉：武汉科技学院,2007.
[33]Texas Instruments, Code Composer Studio User's Guide. February 2000.
[34]Texas Instruments, TMS320DSP/BIOS User's Guide. November 2002.
[35]Texas Instruments, TMS320C6000 Chip Support Library API Reference Guide. May 2004.
[36]http://wenku.baidu.com/view/5e6864030740be1e650e9a6f.html.
[37]Texas Instruments, TVP5150A/VP5150AM1 Ultralow Power NTSC/PAL/SECAM Video Decoder With Robust Sync Decoder Data Manual. May 2004.
[38]Texas Instruments, SAA7120H;SAA7121H Digital video encoder Product. Specific-ation Oct 2002.
[39]Texas Instruments, TMS320C6000 Assembly Language Tools User's Guide. April 2004.
[40]Texas Instruments, Real-Time Digital Video Transfer via High-Speed RTDX. May 2002.
[41]彭启宗等.DSP集成开发环境：CCS及DSP/BIOS的原理与应用.北京：电子工业出版社,2004.
[42]http://wenku.baidu.com/view/264eca6925c52cc58bd6be31.html
[43]Rafael C, Gonzalez, Richard E. Woods著.阮秋琦等译.数字图像处理(第二版).北京：电子工业出版社,2003.
[44]David J. Heeger, James R. Bergen. Pyramid-based texture analysis/synthesis. Annual Conference on Computer Graphics. Los Angeles,1995:229-238.
[2]林学言,王宏等.计算机视觉：一种现代的方法.北京：电子工业出版社,2004.
[3]刘瑞祯,于仕琪. OpenCV教程——基础篇.北京：北京航空航天人学出版社,2007.
[4]http://www.opencv.org.cn/forum/viewforum.php?f=9
[5]吴冬梅,张玉杰.DSP技术及应用.北京：北京大学出版社,2006.
[6]http://www.zhinengjiaotong.com/tech/show-9. html
[7]Leibe B, Leonardis A, Schiele B. Combined object categorization and segmentation with an implicit shape model. European Conference on Computer Vision, Switzerland, 2004:3-10.
[8]Felzenszwalb P, McAllester D, Ramanan D. A Discriminatively Trained, Multiscale, Deformable Part Mode. CVPR, USA,2008:1-8.
[9]Huang Y, Huang K, et al. Enhanced Biologically Inspired Model. CVPR, USA,2008:1-8.
[10]Papageorgiou C, Poggio T. A Trainable System for Object Detection. International Journal of Computer Vision.2000,38(1):15-33.
[11]Kapoor A, Winn J. Located Hidden Random Fields:Learning Discriminative Parts for Object Detection. European Conference on Computer Vision, Austria,2006:2-13
[12]Furst L, Fidler S, Leonardis A. Selecting Features for Object Detection Using an Adboost compatible Evaluation Function.2008, Pattern Recognition Letters(29): 1603-1612.
[13]J. Fang, G. Qiu. Car/non-car classification in an informative sample subspace. In Proceedings of ICPR, Hong kong,2006:962-965.
[14]Z. Zhu, Y. Zhao, H. Lu. Sequential architecture for efficient car detection. IEEE Conference on Computer Vision and Pattern Recognition, USA,2007:1-8.
[15]http://www.docin.com/p-5289933.html
[16]Haritaoglu I, Harwood D, Davis L. S. Real-time surveillance of people and the activities. IEEE Trans Pattern Anal.2000, Machine Intel(22):809-830.
[17]C. Wren, A. Azarbaygaui. Pfinder:Real-time tracking of the human body. IEEE Tran Pattern Anal.1997, Machine Intel(19):780-785.
[18]A. J. SCHOFIELD, A. MEHTA, T. J. STOXHAM. A system for counting people in video images using neural networks to identify the background scene. Pattern Recognition,1996, 29(8):1421-1428.
[19]谷口庆治.数字图像处理------应用篇.北京：科学出版社,2002.
[20]鲁一杰,赵攀,彭强等.基于DSP的视频车辆实时检测系统的实现.计算机应用研究.2009,26(2)：758-759.
[21]Rainer Lien hart, Jochen Maydt. An Extended Set of Haar-like Features for Rapid Object Detection. IEEE ICIP, New York,2002:900-903.
[22]Paul Viola, Michael Jones. Robust Real-time Object Detection. IEEE International Conference on Computer Vision, Canada,2001:1-22.
[23]唐立群,郭庆昌,李永华.数字图像模式识别方法分析.哈尔滨：哈尔滨工程大学出版社,2008.
[24]Freund, Schapire. A decision-theoretic generalization of on-line learning and an application to boosting. In Machine Learning:Proceedings of the Thirteenth International Conference, SanFrancisco,1996:148-156.
[25]Robert E, Yoram Singer. Improved Boosting Algorithms Using Confidence-rated Predictions. Machine Learning, Netherlands,1999:297-336.
[26]Jerome Friedman, Trevor Hastie, Robert Tibshirani. Additive Logistic Regression: a Statistical View of Boosting. Annals of Statistics,2000,28(2):337-407.
[27]Lior W, Lan M. Robust boosting for learning from few example [J]. IEEE Conference on Computer Vision and Pattern Recognition, San Diego,2005:395-364.
[28]Chuck White等著.周生炳等译.XML从入门到精通(黄金版).北京：电子工业出版社,2002.
[29]王跃宗,刘京会.TMS320DM642DSP应用系统设计与开发.北京：人民邮电出版社,2009.
[30]李芳慧,王飞,何佩琨.TM320C6000系列DSP原理与应用(第2版)[M].北京：电子工业出版社,2003.
[31]隋忠磊.基于DM642的数字图像采集处理系统[D].内蒙古：内蒙古工业大学,2007.
[32]李保民.基于DM642的视频图像实时处理模块的设计与实现[D].武汉：武汉科技学院,2007.
[33]Texas Instruments, Code Composer Studio User's Guide. February 2000.
[34]Texas Instruments, TMS320DSP/BIOS User's Guide. November 2002.
[35]Texas Instruments, TMS320C6000 Chip Support Library API Reference Guide. May 2004.
[36]http://wenku.baidu.com/view/5e6864030740be1e650e9a6f.html.
[37]Texas Instruments, TVP5150A/VP5150AM1 Ultralow Power NTSC/PAL/SECAM Video Decoder With Robust Sync Decoder Data Manual. May 2004.
[38]Texas Instruments, SAA7120H;SAA7121H Digital video encoder Product. Specific-ation Oct 2002.
[39]Texas Instruments, TMS320C6000 Assembly Language Tools User's Guide. April 2004.
[40]Texas Instruments, Real-Time Digital Video Transfer via High-Speed RTDX. May 2002.
[41]彭启宗等.DSP集成开发环境：CCS及DSP/BIOS的原理与应用.北京：电子工业出版社,2004.
[42]http://wenku.baidu.com/view/264eca6925c52cc58bd6be31.html
[43]Rafael C, Gonzalez, Richard E. Woods著.阮秋琦等译.数字图像处理(第二版).北京：电子工业出版社,2003.
[44]David J. Heeger, James R. Bergen. Pyramid-based texture analysis/synthesis. Annual Conference on Computer Graphics. Los Angeles,1995:229-238.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |