车辆遮挡检测的研究与应用
|
摘要
运动车辆检测是智能交通领域应用研究中关键的一环,是获得交通信息的基础和保证。但是检测过程中发生的车辆遮挡严重影响了车辆的统计。为了更好进行运动车辆检测,遮挡检测成为一个必不可少的环节。
本文针对高清图像的车辆检测,首先对图像预处理方法进行了研究,并通过设定感兴趣区域,去除干扰区域减少算法计算量。然后研究了现有的各种前景检测算法,比较各种算法的优劣性,并在此基础上提出了自己的前景检测算法。通过本文的算法,可以较快的从高清图像中准确的提取出车辆,在前景检测算法的基础上,实现更精确的车辆轮廓提取,为接下来车辆遮挡检测打下了很好的基础。
论文的主要工作如下：
(1)图像的获得与预处理方面：摄像头的架接,图像的获取,图像的预处理,比较边缘检测算法,选择出一种能很好提取车辆边缘的边缘检测算法。
(2)前景检测方面：针对高清图像,提出一种在时间和效果上都比较理想的前景检测算法,解决了现有背景建模算法造成的车辆空洞和断层。
(3)车辆遮挡方面：在前景检测的基础上展开车辆遮挡检测研究,对于部分车辆遮挡,利用凸包计算出是否发生遮挡,并根据凸包差来对遮挡车辆进行分割。对于严重遮挡时车辆信息不全的特点,计算获得车辆角点,通过计算角点的光流来判断车辆是否发生遮挡,并根据相同运动向量,划分角点区域来处理车辆遮挡。
(4)软件系统方面：根据前面的算法研究,设计车辆图像遮挡检测的整个流程。设计各个模块、算法流程以及具体的实现方法,以便更加直观的体现出算法的效果。
通过大量高清图像进行测试,比较多种算法效果,从而不断对本文算法进行改进。最终结果表明新的前景检测算法计算量小而且效果好,遮挡检测算法能快速、有效的检测出车辆遮挡并处理。
It is the most important part to extract moving vehicles in application of intelligent transportation field, which is the basis and guarantee to obtain traffic information. But the vehicle occlusion occurred during detection seriously affect the vehicle's statistics. In order to get the moving vehicle detection, vehicle occlusion detection is as an essential link.
In vehicles detection of high-definition images, first of all, this article study on the image pre-processed method, and remove interference region to reduce calculated quantities by setting regions of interest. Then compare the advantages and disadvantages of various algorithms after the study of various existing detection algorithms. And put forward a method for prospect detection. Using this algorithm, you can extract vehicles fast and accurately from high definition images and achieve a more accurate contour extraction on the basis of prospect detection, which lay a good foundation for the next vehicle occlusion detection.
The main work in this thesis is as follows:
(1) Image acquisition and preprocessing: the received of camera, image acquisition, image preprocessing, compare of edge detection algorithms, the chosen of a good edge detection algorithm to extract the edge of vehicles.
(2) Prospect detection: propose a method for prospect detection in high-definition image which is ideal both in time and effect, it can solve the empty and chasm caused by the existing background modeling algorithms background modeling algorithm.
(3) Vehicle occlusion: use the convex hull to decide whether occlusion occurred or not in partial occlusion on the basis of prospect detection, and divide occlusion vehicles by the difference of convex hull. For serious occlusion, vehicle information is incomplete, we obtain vehicle comers by calculation, and calculate the optical flow of comer points to determine the occurrence of occlusion, then according to the same motion vector, handle vehicle occlusion by dividing the region of comer.
(4) Software system: according to the previous algorithm, design the entire process of vehicle occlusion detection. Design every module, the algorithm's process and specific method of implementation, which can reflect a more intuitive algorithm result.
We continuously improve this algorithm by testing a large number of high-definition images and comparing the effect of variety of algorithms. Final results show that the new prospect detection algorithm has the virtue of less computation and good effect. Occlusion detection algorithm can detect and deal with vehicle occlusion quickly and effectively.
本文针对高清图像的车辆检测,首先对图像预处理方法进行了研究,并通过设定感兴趣区域,去除干扰区域减少算法计算量。然后研究了现有的各种前景检测算法,比较各种算法的优劣性,并在此基础上提出了自己的前景检测算法。通过本文的算法,可以较快的从高清图像中准确的提取出车辆,在前景检测算法的基础上,实现更精确的车辆轮廓提取,为接下来车辆遮挡检测打下了很好的基础。
论文的主要工作如下：
(1)图像的获得与预处理方面：摄像头的架接,图像的获取,图像的预处理,比较边缘检测算法,选择出一种能很好提取车辆边缘的边缘检测算法。
(2)前景检测方面：针对高清图像,提出一种在时间和效果上都比较理想的前景检测算法,解决了现有背景建模算法造成的车辆空洞和断层。
(3)车辆遮挡方面：在前景检测的基础上展开车辆遮挡检测研究,对于部分车辆遮挡,利用凸包计算出是否发生遮挡,并根据凸包差来对遮挡车辆进行分割。对于严重遮挡时车辆信息不全的特点,计算获得车辆角点,通过计算角点的光流来判断车辆是否发生遮挡,并根据相同运动向量,划分角点区域来处理车辆遮挡。
(4)软件系统方面：根据前面的算法研究,设计车辆图像遮挡检测的整个流程。设计各个模块、算法流程以及具体的实现方法,以便更加直观的体现出算法的效果。
通过大量高清图像进行测试,比较多种算法效果,从而不断对本文算法进行改进。最终结果表明新的前景检测算法计算量小而且效果好,遮挡检测算法能快速、有效的检测出车辆遮挡并处理。
It is the most important part to extract moving vehicles in application of intelligent transportation field, which is the basis and guarantee to obtain traffic information. But the vehicle occlusion occurred during detection seriously affect the vehicle's statistics. In order to get the moving vehicle detection, vehicle occlusion detection is as an essential link.
In vehicles detection of high-definition images, first of all, this article study on the image pre-processed method, and remove interference region to reduce calculated quantities by setting regions of interest. Then compare the advantages and disadvantages of various algorithms after the study of various existing detection algorithms. And put forward a method for prospect detection. Using this algorithm, you can extract vehicles fast and accurately from high definition images and achieve a more accurate contour extraction on the basis of prospect detection, which lay a good foundation for the next vehicle occlusion detection.
The main work in this thesis is as follows:
(1) Image acquisition and preprocessing: the received of camera, image acquisition, image preprocessing, compare of edge detection algorithms, the chosen of a good edge detection algorithm to extract the edge of vehicles.
(2) Prospect detection: propose a method for prospect detection in high-definition image which is ideal both in time and effect, it can solve the empty and chasm caused by the existing background modeling algorithms background modeling algorithm.
(3) Vehicle occlusion: use the convex hull to decide whether occlusion occurred or not in partial occlusion on the basis of prospect detection, and divide occlusion vehicles by the difference of convex hull. For serious occlusion, vehicle information is incomplete, we obtain vehicle comers by calculation, and calculate the optical flow of comer points to determine the occurrence of occlusion, then according to the same motion vector, handle vehicle occlusion by dividing the region of comer.
(4) Software system: according to the previous algorithm, design the entire process of vehicle occlusion detection. Design every module, the algorithm's process and specific method of implementation, which can reflect a more intuitive algorithm result.
We continuously improve this algorithm by testing a large number of high-definition images and comparing the effect of variety of algorithms. Final results show that the new prospect detection algorithm has the virtue of less computation and good effect. Occlusion detection algorithm can detect and deal with vehicle occlusion quickly and effectively.
引文
[1]张智文.中国ITS发展战略构想.中国机电工业,2000年09期.
[2]李峰.智能交通系统在国外的发展趋势.国外公路,1999,(01).
[3]智能交通系统.国家智能交通系统工程技术研究中心网站：http://www.itsc.com.cn/
[4]王涛.检阅中国ITS——写在北京奥运十大智能交通管理系统全面启动之际.中国交通信息产业.China ITS Journal,2008年07期
[5]Nam T Nguyen, SvethaVenkatesh, Geoff West, Hung H. Bui.监控系统中的多摄像机协同(英文).自动化学报,2003,(03).
[6]Hieu T.Nguyen, MarCel Worring, Rein van den Boomgaard. Occlusion robust adaptive template tracking.Proceedings of Eighth IEEE International Conference Computer Vision,Vancouver, Canada,2001:678-683.
[7]Fangling Chang,Li Ma,Yiz heng Qiao.Target Tracking Under Occlusion by Combining Integral-Intensity-Matching with Multi- block-voting.Lecture Notes in Computer Science.2005; 3644(l):77-86.
[8]Franco Oberti, Simona Calcagno, Michela Zara and Calo S.Regazzoni. Robust tracking of humans and vehicles in cluttered scenes with occlusions. IEEE International Conference on Image Processing, Rochester, New York, USA,2002: 629-632.
[9]Shunsuke Kamijo, Yasuyuki Matsushita, Katsushi Ikeuchi, Masao Sakauchi.
Occlusion robust tracking utilizing spatio-temporal markup random
fieldmodel[A].Proceedings of 15th International Conference on Pattern Recognition
[C], Barcelona, Spain,2000:140-144.
[10]Ying Wn,Ting Yu, Guang Hua. Tracking appearances with occlusions. IEEE
Computer Society Conference on Computer Vision and Pattern Recognition,Madison,
Wisconsin USA 2003:789-795.
[11]K.Nummiaro, E.Koller-Meier, L.Van Gool. An adaptive color-based particle
filter. Image and Vision Computing.2003,21(1):99-110.
[12]Yucel Altunbasak, A.Murat Tekalp. Occlusion-adaptive, content-based mesh
design and forward tracking. IEEE Transactions on Image Processing.1997,6
(9):1270-1280.
[13]Jian-Wei Zhao, Peng Wang, Cong-Qing Liu. An object tracking algorithm based on occlusion mesh model. International Conference on Machine Learning and Cybernetics, Beijing, China,2002:288-292.
[14]OpenCV中文网站http://www.opencv.org.cn/index.php/
[15]章毓晋.图象处理和分析.北京：清华大学出版社,1999.
[16]王继芳.数字图象数学形态学算法研究与应用.解放军信息工程大学,2005.
[17]Rafael C.Gonzalez. Richard E.Woods. Digital Image ProcessingSecond Edition.Publishing House of Electronics Industry,2007.
[18]Canny, J., A Computational Approach To Edge Detection, IEEE Trans. Pattern Analysis and Machine Intelligence,1986,679-714.
[19]王亮,胡卫明,谭铁牛.人运动的视觉分析综述.计算机学报25卷,第3期2002.
[20]Stauffer, C, Grimson, WEL. Adaptive background mixture models for real-time tracking. Computer Vision and Pattern Recognition,1999,2(6),23-25.
[21]Lee D-S. On line Adaptive Gaussian Mixture Learning for Video Applications. Statistical Methods in Video Processing, LNCS3247.2004,3247:105-116.
[22]Power PW, Schoonees JA. Understanding background mixture models for foregrounds segmentation. Proceedings of Image and Vision Computing. New Zealand: Auckland,2002:267-271.
[23]向海霞,王卫星.车辆检测中的自适应背景更新算法研究.计算机工程与应用,2009.
[24]高岚,董慧颖,兰利宝.自适应背景下运动目标阴影检测算法研究.现代电子技术,2008.
[25]陈兵旗,孙明.实用数字图像处理与分析.清华大学出版社,2008.
[26]苏志同,肖建军,熊昌镇.基于边缘信息的高清图像车辆检测算法.2009年中国高校通信类院系学术研讨会(IST检索),2009.
[27]裘镇宇,危辉.基于Freeman的边缘跟踪算法及直线段检测.研究与设计,2008年24卷1期.
[28]张正峰,马少飞.新的种子点区域填充算法.计算机工程与应用,2009,45(6),201.
[29]TaoYnag, Satn Z.Li, QuanPan, Jing Li. Real-time mutipile objeets tracking with occlusion Handling in dynamic scenes.Computer Vision and Pattern Recognition, San Deigo, CA, USA,2005:970-975.
[30]陈平,江国昭.基于有序点列的平面点集凸包的新算法.科技通报,2007年23卷6期.
[31]夏侯玉娇,龚声荣.结合Gaussian分布和LK光流法的视频对象分割算法.微电子学与计算机,2009年26卷6期.
[32]Douglas DeCarlo, Dimitris Metaxas. Optical flow constraints on deform able models with applications to face tracking. International journal of computer vision.2000,38(2):99-127.
[33]Berthold K.P. Horn and Brian G Schunck. Determining Optical Flow. Artificial Intelligence,17, pp.185-203,1981.
[34]Lucas, B., and Kanade, T. An Iterative Image Registration Technique with an Application to Stereo Vision, Proc. of 7th International Joint Conference on Artificial Intelligence,674-679.
[35]赵文彬,张艳宁.角点检测技术综述.计算机应用研究,2006,10：16-19.
[36]沈美丽,宋华军.基于角点检测的光流目标跟踪算法.电子器件,2007,30(4)：1397—1399.
[37]张凯.视频运动检测算法的研究和分析.2007,Vol,27, No.1:26-28.
[38]朱克忠.一种运动目标检测与跟踪快速算法的研究.科技资讯,2007,NO.20：254—255.
[39]田彦山.基于山峰聚类的聚类上限确定方法.江西师范大学学报,2007年31卷2期.
[2]李峰.智能交通系统在国外的发展趋势.国外公路,1999,(01).
[3]智能交通系统.国家智能交通系统工程技术研究中心网站：http://www.itsc.com.cn/
[4]王涛.检阅中国ITS——写在北京奥运十大智能交通管理系统全面启动之际.中国交通信息产业.China ITS Journal,2008年07期
[5]Nam T Nguyen, SvethaVenkatesh, Geoff West, Hung H. Bui.监控系统中的多摄像机协同(英文).自动化学报,2003,(03).
[6]Hieu T.Nguyen, MarCel Worring, Rein van den Boomgaard. Occlusion robust adaptive template tracking.Proceedings of Eighth IEEE International Conference Computer Vision,Vancouver, Canada,2001:678-683.
[7]Fangling Chang,Li Ma,Yiz heng Qiao.Target Tracking Under Occlusion by Combining Integral-Intensity-Matching with Multi- block-voting.Lecture Notes in Computer Science.2005; 3644(l):77-86.
[8]Franco Oberti, Simona Calcagno, Michela Zara and Calo S.Regazzoni. Robust tracking of humans and vehicles in cluttered scenes with occlusions. IEEE International Conference on Image Processing, Rochester, New York, USA,2002: 629-632.
[9]Shunsuke Kamijo, Yasuyuki Matsushita, Katsushi Ikeuchi, Masao Sakauchi.
Occlusion robust tracking utilizing spatio-temporal markup random
fieldmodel[A].Proceedings of 15th International Conference on Pattern Recognition
[C], Barcelona, Spain,2000:140-144.
[10]Ying Wn,Ting Yu, Guang Hua. Tracking appearances with occlusions. IEEE
Computer Society Conference on Computer Vision and Pattern Recognition,Madison,
Wisconsin USA 2003:789-795.
[11]K.Nummiaro, E.Koller-Meier, L.Van Gool. An adaptive color-based particle
filter. Image and Vision Computing.2003,21(1):99-110.
[12]Yucel Altunbasak, A.Murat Tekalp. Occlusion-adaptive, content-based mesh
design and forward tracking. IEEE Transactions on Image Processing.1997,6
(9):1270-1280.
[13]Jian-Wei Zhao, Peng Wang, Cong-Qing Liu. An object tracking algorithm based on occlusion mesh model. International Conference on Machine Learning and Cybernetics, Beijing, China,2002:288-292.
[14]OpenCV中文网站http://www.opencv.org.cn/index.php/
[15]章毓晋.图象处理和分析.北京：清华大学出版社,1999.
[16]王继芳.数字图象数学形态学算法研究与应用.解放军信息工程大学,2005.
[17]Rafael C.Gonzalez. Richard E.Woods. Digital Image ProcessingSecond Edition.Publishing House of Electronics Industry,2007.
[18]Canny, J., A Computational Approach To Edge Detection, IEEE Trans. Pattern Analysis and Machine Intelligence,1986,679-714.
[19]王亮,胡卫明,谭铁牛.人运动的视觉分析综述.计算机学报25卷,第3期2002.
[20]Stauffer, C, Grimson, WEL. Adaptive background mixture models for real-time tracking. Computer Vision and Pattern Recognition,1999,2(6),23-25.
[21]Lee D-S. On line Adaptive Gaussian Mixture Learning for Video Applications. Statistical Methods in Video Processing, LNCS3247.2004,3247:105-116.
[22]Power PW, Schoonees JA. Understanding background mixture models for foregrounds segmentation. Proceedings of Image and Vision Computing. New Zealand: Auckland,2002:267-271.
[23]向海霞,王卫星.车辆检测中的自适应背景更新算法研究.计算机工程与应用,2009.
[24]高岚,董慧颖,兰利宝.自适应背景下运动目标阴影检测算法研究.现代电子技术,2008.
[25]陈兵旗,孙明.实用数字图像处理与分析.清华大学出版社,2008.
[26]苏志同,肖建军,熊昌镇.基于边缘信息的高清图像车辆检测算法.2009年中国高校通信类院系学术研讨会(IST检索),2009.
[27]裘镇宇,危辉.基于Freeman的边缘跟踪算法及直线段检测.研究与设计,2008年24卷1期.
[28]张正峰,马少飞.新的种子点区域填充算法.计算机工程与应用,2009,45(6),201.
[29]TaoYnag, Satn Z.Li, QuanPan, Jing Li. Real-time mutipile objeets tracking with occlusion Handling in dynamic scenes.Computer Vision and Pattern Recognition, San Deigo, CA, USA,2005:970-975.
[30]陈平,江国昭.基于有序点列的平面点集凸包的新算法.科技通报,2007年23卷6期.
[31]夏侯玉娇,龚声荣.结合Gaussian分布和LK光流法的视频对象分割算法.微电子学与计算机,2009年26卷6期.
[32]Douglas DeCarlo, Dimitris Metaxas. Optical flow constraints on deform able models with applications to face tracking. International journal of computer vision.2000,38(2):99-127.
[33]Berthold K.P. Horn and Brian G Schunck. Determining Optical Flow. Artificial Intelligence,17, pp.185-203,1981.
[34]Lucas, B., and Kanade, T. An Iterative Image Registration Technique with an Application to Stereo Vision, Proc. of 7th International Joint Conference on Artificial Intelligence,674-679.
[35]赵文彬,张艳宁.角点检测技术综述.计算机应用研究,2006,10：16-19.
[36]沈美丽,宋华军.基于角点检测的光流目标跟踪算法.电子器件,2007,30(4)：1397—1399.
[37]张凯.视频运动检测算法的研究和分析.2007,Vol,27, No.1:26-28.
[38]朱克忠.一种运动目标检测与跟踪快速算法的研究.科技资讯,2007,NO.20：254—255.
[39]田彦山.基于山峰聚类的聚类上限确定方法.江西师范大学学报,2007年31卷2期.