机车司机视野扩展系统及路轨障碍物检测的研究
|
摘要
铁路运输在我国交通运输领域一直都占主导地位,在经济社会发展中具有特殊重要的作用。近年来,随着铁路的大面积提速调图,使得铁路的行车安全问题显得更加突出。尤其是,我国将长期以来一直实施的正、副司机操纵列车模式改为单司机值乘模式后,司机很难同时兼顾行车信号确认、列车的操纵及路面状况的瞭望等。
针对这种现状,本文提出了机车司机视野扩展系统的框架及设计方案。虽然目前已有一些成熟的智能公路交通技术可以直接使用于铁路交通领域,但铁路智能监控视频有其自身突出的特点：铁路轨道部分有枕木、碎石道砟,不似公路路面那么干净；背景复杂,且在行车过程当中随时会改变等。本系统旨在通过利用铁路路轨特点实时的分析火车视频监控图像序列,检测、跟踪并分析威胁火车安全行驶的路障目标,达到减少事故发生率、提高铁路运输安全的目的。
系统整体设计思路是将铁路静态路轨障碍物的检测和智能视频监控序列中动态障碍物的检测、跟踪相分开来。对于静态路轨障碍物的检测,主要运用了特征提取和特征匹配的算法。该算法将铁路障碍物的检测范围限制在图像铁轨部分,对检测窗内图像的纹理和灰度特征进行分析,从而判定障碍物的存在与否。而动态障碍物的检测是通过对运动目标的提取、运动目标的跟踪和目标的运动轨迹分析等部分展开。用光流背景建模方法提取运动目标,而后用Kalman+Mean shift算法对其进行实时跟踪,在跟踪的过程当中分析目标的运动轨迹。
论文按系统框架、算法设计及系统仿真的顺序对机车司机视野扩展系统进行详细介绍。首先,提出机车司机视野扩展系统的总体设计方案和框架；而后,对系统算法原理进行阐述；最后,用仿真环境下采集的视频图像序列和图片对算法进行测试。实验结果表明,本系统的运行速度较快且抗噪能力较强,有一定的实用价值。
Railway transport is the most widely used transportation method in China. Focusing on the increasing number of the railway passengers, the Ministry of Railways has been improving the speed of train for six times. High-speed train runs at much faster speeds than traditional train, so the traditional detection method with human-vision is not suitable for road surface condition observation of the high-speed train under the Single-Driver Duty mode. In order to deal with these problems, people begin to develop automatic monitoring system to replace the human to complete these work.
Although there are some mature intelligent video surveillance systems can be used directly in the field of railway transport, railway intelligent surveillance video has its own characteristics such as railway is not as clean as the road surface cause of the railway sleepers and ballast. Therefore, this paper describes development of the locomotive driver's vision expansion system.
The system consists of two parts:movable obstacle detection and static obstacle detection. The static roadblock detection algorithm improves the detection speed by limiting the scope of detection in the tracks area. After sets up the Obstacle detection window, searches the roadblock target by analyzing texture feature and gray feature of the picture within the detection window. The most important part of movable roadblock detection is target tracking. Firstly, use the Inter-frame Difference algorithm to extract the moving region; secondly, eliminate the moving background pixels by establishing optical flow background model and get foreground objects; thirdly, track these objects and record the motion curve; finally, analyze the motion curve.
After making sure the existence of dangerous obstacles, send alerts to driver. The development of system is aimed at reduce accidents and improve railway transport safety by detect, track and analyze the roadblock target in the real-time video surveillance image sequence of train. The experiments show that this study has practical value in detect、track and analyze roadblock object.
针对这种现状,本文提出了机车司机视野扩展系统的框架及设计方案。虽然目前已有一些成熟的智能公路交通技术可以直接使用于铁路交通领域,但铁路智能监控视频有其自身突出的特点：铁路轨道部分有枕木、碎石道砟,不似公路路面那么干净；背景复杂,且在行车过程当中随时会改变等。本系统旨在通过利用铁路路轨特点实时的分析火车视频监控图像序列,检测、跟踪并分析威胁火车安全行驶的路障目标,达到减少事故发生率、提高铁路运输安全的目的。
系统整体设计思路是将铁路静态路轨障碍物的检测和智能视频监控序列中动态障碍物的检测、跟踪相分开来。对于静态路轨障碍物的检测,主要运用了特征提取和特征匹配的算法。该算法将铁路障碍物的检测范围限制在图像铁轨部分,对检测窗内图像的纹理和灰度特征进行分析,从而判定障碍物的存在与否。而动态障碍物的检测是通过对运动目标的提取、运动目标的跟踪和目标的运动轨迹分析等部分展开。用光流背景建模方法提取运动目标,而后用Kalman+Mean shift算法对其进行实时跟踪,在跟踪的过程当中分析目标的运动轨迹。
论文按系统框架、算法设计及系统仿真的顺序对机车司机视野扩展系统进行详细介绍。首先,提出机车司机视野扩展系统的总体设计方案和框架；而后,对系统算法原理进行阐述；最后,用仿真环境下采集的视频图像序列和图片对算法进行测试。实验结果表明,本系统的运行速度较快且抗噪能力较强,有一定的实用价值。
Railway transport is the most widely used transportation method in China. Focusing on the increasing number of the railway passengers, the Ministry of Railways has been improving the speed of train for six times. High-speed train runs at much faster speeds than traditional train, so the traditional detection method with human-vision is not suitable for road surface condition observation of the high-speed train under the Single-Driver Duty mode. In order to deal with these problems, people begin to develop automatic monitoring system to replace the human to complete these work.
Although there are some mature intelligent video surveillance systems can be used directly in the field of railway transport, railway intelligent surveillance video has its own characteristics such as railway is not as clean as the road surface cause of the railway sleepers and ballast. Therefore, this paper describes development of the locomotive driver's vision expansion system.
The system consists of two parts:movable obstacle detection and static obstacle detection. The static roadblock detection algorithm improves the detection speed by limiting the scope of detection in the tracks area. After sets up the Obstacle detection window, searches the roadblock target by analyzing texture feature and gray feature of the picture within the detection window. The most important part of movable roadblock detection is target tracking. Firstly, use the Inter-frame Difference algorithm to extract the moving region; secondly, eliminate the moving background pixels by establishing optical flow background model and get foreground objects; thirdly, track these objects and record the motion curve; finally, analyze the motion curve.
After making sure the existence of dangerous obstacles, send alerts to driver. The development of system is aimed at reduce accidents and improve railway transport safety by detect, track and analyze the roadblock target in the real-time video surveillance image sequence of train. The experiments show that this study has practical value in detect、track and analyze roadblock object.
引文
[1]曾青中.铁路机车单司机值乘的问题与对策[J].中国科技信息,2006,(24)：76-77.
[2]商静.道路障碍物检测的研究[D].长春理工大学,2010.
[3]李庆忠,陈显华,顾伟康等.基于彩色立体视觉的障碍物快速检测方法[J].计算机科学,2003,30(9)：72-75.
[4]阮秋琦.数字图像处理学[M].第五版.北京：电子工业出版社,2004.
[5]段瑞玲,李庆祥,李玉和等.图像边缘检测方法研究综述[J].光学技术,2005,31(3)：415-419.
[6]宋娟.路轨自动检测系统及障碍物识别技术的研究[D].浙江大学信息学院,2008.
[7]张恒博,欧宗瑛.一种基于颜色基元共生矩阵的图像检索方法[J].计算机工程,2007,33(14)：171-173.
[8]张云彬,张永生.基于图像纹理特征的目标快速检索[J].高技术通讯,2004,14(8)：11-14.
[9]郭忠伟,郑华利,马仁安等.一种基于纹理特征的战场目标图像挖掘方法[J].舰船电子工程,2010,30(4)：65-67,122.
[10]王孝艳,张艳珠,董慧颖,李媛等.运动目标检测的三帧差算法研究[J].沈阳理工大学学报,2011,30(6)：82-85.
[11]张水发,张文生,丁欢等.融合光流速度与背景建模的目标检测方法[J].中国图像图形学报,2011,16(2)：236-243
[12]荣太,吴元昊,王明佳等.视频目标跟踪算法综述[J].视频应用与工程,2010,34(12)：135-138.
[13]丁一,毛征,余欢.基于对比度的目标跟踪系统及算法研究[J].中国科技论文在线
[14]Perez M M, Dennis T J. An adaptive implementation of the SUSAN method for image edge and feature detection[C]. Proceedings of International Conference on Image Processing. Santa Barbara, CA,1997:394-397.
[15]马桂珍,房宗良,姚宗中SUSAN边缘检测算法性能分析与比较[J].现代电子技术,2007,8：189-191.
[16]Xiang Zhang, Yuan-Ming Dai,Zhang-Wei Chen, Huai-Xiang Zhang. An improved Mean Shift tracking algorithm based on color and texture feature[C]. International Conference on Wavelet Analysis and Pattern Recognition (ICWAPR). Qingdao,China,2010:38-43.
[17]Mirabi M, Javadi S. People Tracking in Outdoor Environment Using Kalman Filter[C]. Third International Conference on Intelligent Systems, Modelling and Simulation (ISMS). Kota Kinabalu,2012:303-307.
[18]董宏辉,葛大伟,秦勇,等.基于智能视频分析的铁路入侵检测技术研究[J].中国铁道科学,2010,31(2)：121-124
[19]彭飞,陈维荣,冒波波,等.基于Canny边缘检测和聚合接续法的路轨边缘提取方法[J].铁道学报,2012,34(2)：52-57.
[20]柴世红.基于边缘检测的铁轨识别[J].铁路计算机应用,2009,18(4)：1-3.
[21]Xiangjian He, Jianmin Li, Daming Wei, Wenjing Jia, Qiang Wu. Canny edge detection on a virtual hexagonal image structure[C]. Joint Conferences on Pervasive Computing (JCPC). Tamsui, Taipei, 2009:167-172.
[22]Demigny D, Lorca F G, Kessal L. Evaluation of edge detectors performances with a discrete expression of Canny's criteria[C]. Proceedings of the International Conference on Image Processing. Washington, DC, 1995:169-172.
[23]关鹏,顾晓东,张立明等.一种基于图像处理的铁轨自动检测方法[J].计算机工程,2007,33(19)：207-209,212.
[24]Adams R, Bischof L. Seeded region growing [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1994, 16(6):641-647.
[25]刘丽,匡纲要.图像纹理特征提取方法综述[J].中国图形图像学报A,2009,14(4)：622-635.
[26]杨凯陟,程英蕾.基于灰度共生矩的SAR图像纹理特征提取方法[J].电子科技,2011,24(11)：66-69.
[27]Yu Jian. Texture Image Segmentation Based on Gaussian Mixture Models and Gray Level Co-occurrence Matrix [C]. International Symposium on Information Science and Engineering (ISISE).Shanghai, China, 24-26 Dec.2010:149-152.
[28]Junqiu Wang, Yasushi Yagi. Integrating Color and Shape-Texture Features for Adaptive Real-Time Object Tracking [J].IEEE Transactions on Image Processing, 2008,17(2):235-240.
[29]危自福,毕笃彦,杨俭.基于多级纹理特征和Mean-Shift的灰度目标跟踪[J].计算机应用,2010,30(6)：1568-1572.
[30]曲云腾,李康平,杜秀霞.基于kalman预测的人体运动目标跟踪[J].计算以系统应用,2011,20(1)：137-140.
[31]余静,游志胜.自动目标识别与跟踪技术研究综述[J].计算机应用研究,2005,22(1)：12-15.
[32]常发亮,刘雪,王华杰等.基于均值漂移与卡尔曼滤波的目标跟踪算法[J].计算机工程与应用,2007,43(12)：50-52.
[33]邱泰生.基于均值漂移的视频目标跟踪算法的研究[D].中山大学,2010.
[34]Comaniciu D, Ramesh V, Meer P. Real-time tracking of non-rigid objects using mean shift[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Hilton Head Island, SC,2000:142-149.
[35]Comaniciu D, Ramesh V, Meer, P. Kernel-based object tracking [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003,25(5), 564-577.
[36]孙承志,熊田忠,吉顺平等.基于差分的光流法在目标检测跟踪中的应用[J].机床与液压,2010,38(14)：59-62.
[37]Hongxia Chu, Wang, K. Target tracking based on mean shift and improved kalman filtering algorithm[C]. IEEE International Conference on Automation and Logistics (ICAL). Shenyang, China,2009:808-812.
[38]Tai J, Tsang S, Lin C, Song K. Real-time image tracking for automatic traffic monitoring and enforcement application [J]. Image and Vision Computing, 2004, 22(6):485-501.
[39]D Comaniciu, P Meer. Mean shift: A robust approach toward feature space analysis [J]. IEEE Transaction on Pattern Analysis and Machine Intelligence (S0162-8828), 2002,24(5):603 - 619.
[40]Feng Shimin, Guan Qing, Xu Sheng,et al. Human Tracking Based on Mean Shift and Kalman Filter[C]. International Conference on Artificial Intelligence and Computational Intelligence (AICI).Shanghai, China, 2009:518-522.
[41]程建,杨杰.一种基于均值移位的红外目标跟踪新方法[J].红外与毫米波学报,2005,24(3),231-235.
[42]程建,周越,蔡念等.基于粒子滤波的红外目标跟踪[J].红外与毫米波学报,2006,25(2),113-117.
[43]李善青,唐亮,刘科研等.一种快速的自适应目标跟踪方法[J].计算机研究与发展,2012,49(2)：383-391.
[44]Yizong Cheng, Mean shift, mode seeking, and clustering [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995,17(8):pp. 790-799.
[45]D. Comaniciu, V. Ramesh, P. Meer. Kernel based object tracking [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003,25(5):564-577.
[46]常发亮,马丽,乔谊正等.视频序列中面向人的多目标跟踪算法[J].控制与决策,2007,22(4)：418-422.
[47]许小勇,钟太勇.三次样条差值函数的构造与Matlab实现[J].自动测量与控制,2006,25(11)：76-78.
[48]齐鹤.运动目标检测及运动轨迹分析[D].大连理工大学,2009.
[49]Milan Sonka, Vaclav Hlavac, Roger Boyle. Image Processing, Analysis, and Machine Vision[M]. Thomson Asia Pte Led,2002.
[2]商静.道路障碍物检测的研究[D].长春理工大学,2010.
[3]李庆忠,陈显华,顾伟康等.基于彩色立体视觉的障碍物快速检测方法[J].计算机科学,2003,30(9)：72-75.
[4]阮秋琦.数字图像处理学[M].第五版.北京：电子工业出版社,2004.
[5]段瑞玲,李庆祥,李玉和等.图像边缘检测方法研究综述[J].光学技术,2005,31(3)：415-419.
[6]宋娟.路轨自动检测系统及障碍物识别技术的研究[D].浙江大学信息学院,2008.
[7]张恒博,欧宗瑛.一种基于颜色基元共生矩阵的图像检索方法[J].计算机工程,2007,33(14)：171-173.
[8]张云彬,张永生.基于图像纹理特征的目标快速检索[J].高技术通讯,2004,14(8)：11-14.
[9]郭忠伟,郑华利,马仁安等.一种基于纹理特征的战场目标图像挖掘方法[J].舰船电子工程,2010,30(4)：65-67,122.
[10]王孝艳,张艳珠,董慧颖,李媛等.运动目标检测的三帧差算法研究[J].沈阳理工大学学报,2011,30(6)：82-85.
[11]张水发,张文生,丁欢等.融合光流速度与背景建模的目标检测方法[J].中国图像图形学报,2011,16(2)：236-243
[12]荣太,吴元昊,王明佳等.视频目标跟踪算法综述[J].视频应用与工程,2010,34(12)：135-138.
[13]丁一,毛征,余欢.基于对比度的目标跟踪系统及算法研究[J].中国科技论文在线
[14]Perez M M, Dennis T J. An adaptive implementation of the SUSAN method for image edge and feature detection[C]. Proceedings of International Conference on Image Processing. Santa Barbara, CA,1997:394-397.
[15]马桂珍,房宗良,姚宗中SUSAN边缘检测算法性能分析与比较[J].现代电子技术,2007,8：189-191.
[16]Xiang Zhang, Yuan-Ming Dai,Zhang-Wei Chen, Huai-Xiang Zhang. An improved Mean Shift tracking algorithm based on color and texture feature[C]. International Conference on Wavelet Analysis and Pattern Recognition (ICWAPR). Qingdao,China,2010:38-43.
[17]Mirabi M, Javadi S. People Tracking in Outdoor Environment Using Kalman Filter[C]. Third International Conference on Intelligent Systems, Modelling and Simulation (ISMS). Kota Kinabalu,2012:303-307.
[18]董宏辉,葛大伟,秦勇,等.基于智能视频分析的铁路入侵检测技术研究[J].中国铁道科学,2010,31(2)：121-124
[19]彭飞,陈维荣,冒波波,等.基于Canny边缘检测和聚合接续法的路轨边缘提取方法[J].铁道学报,2012,34(2)：52-57.
[20]柴世红.基于边缘检测的铁轨识别[J].铁路计算机应用,2009,18(4)：1-3.
[21]Xiangjian He, Jianmin Li, Daming Wei, Wenjing Jia, Qiang Wu. Canny edge detection on a virtual hexagonal image structure[C]. Joint Conferences on Pervasive Computing (JCPC). Tamsui, Taipei, 2009:167-172.
[22]Demigny D, Lorca F G, Kessal L. Evaluation of edge detectors performances with a discrete expression of Canny's criteria[C]. Proceedings of the International Conference on Image Processing. Washington, DC, 1995:169-172.
[23]关鹏,顾晓东,张立明等.一种基于图像处理的铁轨自动检测方法[J].计算机工程,2007,33(19)：207-209,212.
[24]Adams R, Bischof L. Seeded region growing [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1994, 16(6):641-647.
[25]刘丽,匡纲要.图像纹理特征提取方法综述[J].中国图形图像学报A,2009,14(4)：622-635.
[26]杨凯陟,程英蕾.基于灰度共生矩的SAR图像纹理特征提取方法[J].电子科技,2011,24(11)：66-69.
[27]Yu Jian. Texture Image Segmentation Based on Gaussian Mixture Models and Gray Level Co-occurrence Matrix [C]. International Symposium on Information Science and Engineering (ISISE).Shanghai, China, 24-26 Dec.2010:149-152.
[28]Junqiu Wang, Yasushi Yagi. Integrating Color and Shape-Texture Features for Adaptive Real-Time Object Tracking [J].IEEE Transactions on Image Processing, 2008,17(2):235-240.
[29]危自福,毕笃彦,杨俭.基于多级纹理特征和Mean-Shift的灰度目标跟踪[J].计算机应用,2010,30(6)：1568-1572.
[30]曲云腾,李康平,杜秀霞.基于kalman预测的人体运动目标跟踪[J].计算以系统应用,2011,20(1)：137-140.
[31]余静,游志胜.自动目标识别与跟踪技术研究综述[J].计算机应用研究,2005,22(1)：12-15.
[32]常发亮,刘雪,王华杰等.基于均值漂移与卡尔曼滤波的目标跟踪算法[J].计算机工程与应用,2007,43(12)：50-52.
[33]邱泰生.基于均值漂移的视频目标跟踪算法的研究[D].中山大学,2010.
[34]Comaniciu D, Ramesh V, Meer P. Real-time tracking of non-rigid objects using mean shift[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Hilton Head Island, SC,2000:142-149.
[35]Comaniciu D, Ramesh V, Meer, P. Kernel-based object tracking [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003,25(5), 564-577.
[36]孙承志,熊田忠,吉顺平等.基于差分的光流法在目标检测跟踪中的应用[J].机床与液压,2010,38(14)：59-62.
[37]Hongxia Chu, Wang, K. Target tracking based on mean shift and improved kalman filtering algorithm[C]. IEEE International Conference on Automation and Logistics (ICAL). Shenyang, China,2009:808-812.
[38]Tai J, Tsang S, Lin C, Song K. Real-time image tracking for automatic traffic monitoring and enforcement application [J]. Image and Vision Computing, 2004, 22(6):485-501.
[39]D Comaniciu, P Meer. Mean shift: A robust approach toward feature space analysis [J]. IEEE Transaction on Pattern Analysis and Machine Intelligence (S0162-8828), 2002,24(5):603 - 619.
[40]Feng Shimin, Guan Qing, Xu Sheng,et al. Human Tracking Based on Mean Shift and Kalman Filter[C]. International Conference on Artificial Intelligence and Computational Intelligence (AICI).Shanghai, China, 2009:518-522.
[41]程建,杨杰.一种基于均值移位的红外目标跟踪新方法[J].红外与毫米波学报,2005,24(3),231-235.
[42]程建,周越,蔡念等.基于粒子滤波的红外目标跟踪[J].红外与毫米波学报,2006,25(2),113-117.
[43]李善青,唐亮,刘科研等.一种快速的自适应目标跟踪方法[J].计算机研究与发展,2012,49(2)：383-391.
[44]Yizong Cheng, Mean shift, mode seeking, and clustering [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995,17(8):pp. 790-799.
[45]D. Comaniciu, V. Ramesh, P. Meer. Kernel based object tracking [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003,25(5):564-577.
[46]常发亮,马丽,乔谊正等.视频序列中面向人的多目标跟踪算法[J].控制与决策,2007,22(4)：418-422.
[47]许小勇,钟太勇.三次样条差值函数的构造与Matlab实现[J].自动测量与控制,2006,25(11)：76-78.
[48]齐鹤.运动目标检测及运动轨迹分析[D].大连理工大学,2009.
[49]Milan Sonka, Vaclav Hlavac, Roger Boyle. Image Processing, Analysis, and Machine Vision[M]. Thomson Asia Pte Led,2002.