沥青路面车辙检测算法研究
|
摘要
随着我国经济的飞速发展,我国的公路系统也越来越发达,沥青公路占总公路里程的比例也日益增加。我国早期建设的公路已经出现破损,其中路面破损前期80%属于车辙病害,其破坏路容,危害交通安全。因此,在高速公路飞速发展的今天,车辙检测对公路养护工作具有十分重要的意义,是我国公路养护的重要课题。
本课题来源于南京理工大学与江苏省沪宁高速公路股份有限公司的共同研制的N-1型路面状况智能检测车。为了能更高效,快速的检测出车辙,本文对该系统的子系统“基于结构光的车辙自动检测”进行研究。
路面车辙检测系统主要由车载图像采集系统和离线数据处理系统两大模块组成,本文主要对离线数据处理模块进行研究。首先叙述了车辙检测系技术的发展历程和国内外发展现状,其次介绍了车辙的分类及形成机理,并对基于结构光路面车辙检测系统原理进行了详细的论述。在离线数据处理模块中,通过对原始车辙图像进行灰度化、灰度校正、阈值分割、细化等一系列处理后,再对细化后的曲线进行标定,从而计算出车辙深度。在阈值分割阶段,本文利用基于全局阈值分割和P-tile阈值分割的分段阈值分割方法,实验表明该方法可以有效的消除行车线以及光照不均等造成的干扰,提高图像的质量；在细化阶段,通过研究经典细化算法,针对车辙图像的特点以及结构光图像的灰度分布特点,本文采用了一种直接提取中心线的细化算法,该方法复杂度低,处理效果好,运行时间也比较少；在基于模板标定的车辙深度计算环节,本文尝试用双线性插值的方法对原始图像进行插值,提高原始图像的分辨率,从而使标定精确提高到亚像素级。
最后根据原始车辙图像的特点将车辙图像分类,对不同种类的图像采用不同的处理方法,这样可以正确而有效的检测到车辙并获取细化后的车辙图像,从而提高车辙深度测量的精度。
Along with the rapid development of national economy, national highway system is also more and more developed, the proportion which asphalt highway possess highway overall is also increasing. The highway which has been constructed early has appeared damage, the 80% diseases belong to rutting diseases in pavement damage prophase, it imperils the appearance of the pavement. Therefore, with the rapid development of the highway, road maintenance has become extremely more significant, rut detection is an important issue of our national highway maintenance.
This paper comes from the project N-1 Style Intelligent Data Gathering and Processing Vehicle for the Pavement which is researched by Nanjing University of Science and Technology Computer department and Jiangsu modern engineering detection Co. LTD. In order to detect rut diseases more efficiently, fast, the paper researches for sub-project of the project "automatic detection for the rut diseases based on the structure light".
The automatic detection system for rut diseases is consisted of image acquisition system on-board and data processing system off-line. The paper focuses on the research of data processing module off-line. First, the paper introduced the classification of rut diseases and its formation mechanism. Then, it describes the course of the rutting detection technology development and the present situation of its development at home and abroad. And last the theory of the rutting test system for pavement based on the structure light is discussed in detail. In data processing off-line modules, processing the original rut image through Series of treatment gray-scale melt, the gray-level correction, threshold segmentation, thinning etc. and then after the calibration of the thinning curve and thus we can calculate the rut depth. At the threshold segmentation stage, this paper uses the segmental threshold segmentation method based on global threshold segmentation method and P-tile threshold method, the experiment results show that the method can eliminate the interference caused by traffic lanes and uniform illumination effectively, and improve the image quality. At the image thinning stage, through research for several kinds of classic thinning algorithms, this paper propose a new algorithm of center line extraction directly for rut image aimed at the graphic characteristic of the rut image, and the characteristic of the gray-scale division of structure light, this method complexity is low, the treatment effect is good, running time is less. At the calibration based on the template stage, this paper processes the original rut image by using bilinear interpolation method, and attempts to improve the resolution of the original image, in order to make the calibration precision increased to sub-pixel level.
Finally, according to the characteristics of the original rut image, the paper classifies the rut image to four kinds, and disposes different kind of the rut image using different processing method. And then we can detect the rut disease correctly and effectively, and obtain the thinning ruts of image, so as to improve the rut depth measuring precision.
本课题来源于南京理工大学与江苏省沪宁高速公路股份有限公司的共同研制的N-1型路面状况智能检测车。为了能更高效,快速的检测出车辙,本文对该系统的子系统“基于结构光的车辙自动检测”进行研究。
路面车辙检测系统主要由车载图像采集系统和离线数据处理系统两大模块组成,本文主要对离线数据处理模块进行研究。首先叙述了车辙检测系技术的发展历程和国内外发展现状,其次介绍了车辙的分类及形成机理,并对基于结构光路面车辙检测系统原理进行了详细的论述。在离线数据处理模块中,通过对原始车辙图像进行灰度化、灰度校正、阈值分割、细化等一系列处理后,再对细化后的曲线进行标定,从而计算出车辙深度。在阈值分割阶段,本文利用基于全局阈值分割和P-tile阈值分割的分段阈值分割方法,实验表明该方法可以有效的消除行车线以及光照不均等造成的干扰,提高图像的质量；在细化阶段,通过研究经典细化算法,针对车辙图像的特点以及结构光图像的灰度分布特点,本文采用了一种直接提取中心线的细化算法,该方法复杂度低,处理效果好,运行时间也比较少；在基于模板标定的车辙深度计算环节,本文尝试用双线性插值的方法对原始图像进行插值,提高原始图像的分辨率,从而使标定精确提高到亚像素级。
最后根据原始车辙图像的特点将车辙图像分类,对不同种类的图像采用不同的处理方法,这样可以正确而有效的检测到车辙并获取细化后的车辙图像,从而提高车辙深度测量的精度。
Along with the rapid development of national economy, national highway system is also more and more developed, the proportion which asphalt highway possess highway overall is also increasing. The highway which has been constructed early has appeared damage, the 80% diseases belong to rutting diseases in pavement damage prophase, it imperils the appearance of the pavement. Therefore, with the rapid development of the highway, road maintenance has become extremely more significant, rut detection is an important issue of our national highway maintenance.
This paper comes from the project N-1 Style Intelligent Data Gathering and Processing Vehicle for the Pavement which is researched by Nanjing University of Science and Technology Computer department and Jiangsu modern engineering detection Co. LTD. In order to detect rut diseases more efficiently, fast, the paper researches for sub-project of the project "automatic detection for the rut diseases based on the structure light".
The automatic detection system for rut diseases is consisted of image acquisition system on-board and data processing system off-line. The paper focuses on the research of data processing module off-line. First, the paper introduced the classification of rut diseases and its formation mechanism. Then, it describes the course of the rutting detection technology development and the present situation of its development at home and abroad. And last the theory of the rutting test system for pavement based on the structure light is discussed in detail. In data processing off-line modules, processing the original rut image through Series of treatment gray-scale melt, the gray-level correction, threshold segmentation, thinning etc. and then after the calibration of the thinning curve and thus we can calculate the rut depth. At the threshold segmentation stage, this paper uses the segmental threshold segmentation method based on global threshold segmentation method and P-tile threshold method, the experiment results show that the method can eliminate the interference caused by traffic lanes and uniform illumination effectively, and improve the image quality. At the image thinning stage, through research for several kinds of classic thinning algorithms, this paper propose a new algorithm of center line extraction directly for rut image aimed at the graphic characteristic of the rut image, and the characteristic of the gray-scale division of structure light, this method complexity is low, the treatment effect is good, running time is less. At the calibration based on the template stage, this paper processes the original rut image by using bilinear interpolation method, and attempts to improve the resolution of the original image, in order to make the calibration precision increased to sub-pixel level.
Finally, according to the characteristics of the original rut image, the paper classifies the rut image to four kinds, and disposes different kind of the rut image using different processing method. And then we can detect the rut disease correctly and effectively, and obtain the thinning ruts of image, so as to improve the rut depth measuring precision.
引文
[1]张国伍,中国“十二五”交通运输发展战略—“交通7+1论坛”第十六次会议纪实[R].交通运输系统工程与信息,2009,10(5)：1-10
[2]啜二勇,国外路面自动检测系统发展综述[J].交通标准化,2009,9(204)：96-99
[3]胡霞光,国内外路面快速检测技术的现状与发展[J].中外公路,2003.23(6)：95-99
[4]Chuo, Er-yong. Development Summary of International Pavement Surface Distress Automatic survey System [J]. Transport Standardization,2009(204):96-99
[5]S Rasmussen, J A Krarup. Non-contact Deflection Measurement at High Speed [C] Lisbon, Portugal:The 6th International Conference on the Bearing Capacity of Roads, Railways and Airfields,2002
[6]Kaseko M S, Lo Z P, Ritchie S G. Comparison of Traditional and Neural Classifiers for pavement crack detection. Journal of Transportation Engineering,1993,119(3):402-418
[7]C.F Chen, Chun Jung Hsu. Development of An Automated Airport Pavement Image Collection System [J]. Journal of Marine Science and Technology,2002,10(1):1-7
[8]啜二勇,国内路面自动检测系统研究历程及展望[J].中国高新技术企业杂志,2009(19)：195-196
[9]张围梁,叶中辰.路面无损检测技术的现状与发展[J].吉林交通科技,2007(3)：60-62
[10]李强.高速公路路面车辙检测、评价与预测技术研究[D].东南大学,2007
[11]Koutsopoulos H N, Downey A B. Primitive-based classification of pavement cracking images [J]. Journal of Transportation Engineering,1993,119(3):402-418
[12]田恩杰,高等级公路路面病害自动检测方法研究[D].吉林大学,2007
[13]申爱琴,庄传仪.山区高速公路沥青路面车辙成因与防治[J].筑路机械与施工机械化,2007(6)：1-4
[14]刘钰,沥青路面车辙深度检测系统研究[D].武汉理工大学,2007
[15]汪恩军,公路车辙自动检测系统研制[D].武汉理工大学,2007
[16]王鑫,唐振民.一种新的自动路面车辙检测方法[J].计算机工程与应用,2008,44(10)：246-248
[17]姚运仕,赵万芹,宋宏勋.沥青路面车辙检测技术的探讨[J].筑路机械与施工机械化,2007(6)：8-10
[18]Mallela R, Wang H. Harmonisiong Automated Rut Depth Measurements:Stage 2, Research Report 277 [R]. Land Transport New Zealand,2006
[19]温智源,贾圣东.基于点激光与线结构光的车辙仪对比[J].山西建筑,2010,36(3)： 350-351
[20]吴海艳,路面车辙特征提取与技术方法研究[D].哈尔滨工业大学.2007
[21]曹冰莹,激光路面检测技术集成与开发研究[D].长安大学,2006
[22]徐卫东,严凯,王鑫.基于结构光的路面车辙检测系统[J].公路交通科技·技术版,2009(6)：49-50,60
[23]Hu GongZhu, Gorge Stonman.3D surface solution using structured light and constraint propagation. IEEE-PAMI,1995,11(4):225-231
[24]任明武,高建贞,杨静宇.一种快速实用的灰度校正算法[J].中国图象图形学报,2002,7(6)：548-552
[25]James D F, Robert A S. Spatially variant contrast enhancement using local ranger modification [J]. SPIE,1983,22(3):378-381
[26]王明辉,沥青路面检测中的图像处理技术研究[D].武汉理工大学,2006.5
[27]周敬,图像分割中阈值法的研究[J].机电技术,2010(1)：39-41,65
[28]高建贞,任明武,唐振民.路面裂缝的自动检测与识别[J].计算机工程,2003,29(2)：49-150
[29]Sun T, Neuvo Y. Detail-preserving median based filters in image processing [J].Pattern Recognition Letters,1994,15(4):341-347.
[30]Wang Z, Zhang D. Progressive switching median filter for the removal of impulse noise from highly corrupted images[J].IEEE Trans on Circuits and Systems Ⅱ:Analog and Digital Signal Processing,1999,46(l):78-80.
[31]XING Cang-ju, WANG Shou-jue, DENG Hao-jiang. A new filtering algorithm based on extremum and median value[J]. Journal of image and Graphics,2001,6(6):533-536.
[32]宋琼琼,贾振红.基于人眼视觉特性的自适应中值滤波算法[J].光电子··激光2009,19(1)：128-130
[33]Pok G, Liu J C, Nair A S, Selective removal of impulse noise based on homogeneity level information[J]. IEEE Trans on Image Processing,2003,12(1):85-92
[34]XU Xiao-in, Miller E 1, CHEN Dong-bin, et al. Adaptive two-pass rank order filer to remove impulse noise in highly corrupted images [J]. IEEE Signal Process Letters,2006, 13(7):413-416
[35]WANG Hong-mei, Li Yan-jun, ZHANG Ke. An improved image filtering method for salt and pepper noise [J]. Journal of Optoelectronics · Laser,2007,18(1):113-116.
[36]赵磊,陈琼,陈中.一种新的改进OPTA细化算法[J].计算机应用,2008,10(10),2638.2642
[37]H. Tamura, A comparison of line thinning algorithms from digital geometry viewpoint [J], in Int. Joint conf. Recognition,1978:715-719
[38]A. Datta and S.K.Parui, A Robust Parrallel Thinning Algorithm for Binary Image [J], Pattern Recognition,1994,27(9):1181-1192
[39]冯星奎,李林艳,颜祖泉.一种新的指纹图像细化算法[J].中国图像图形学报,1999,4(10)：835-838
[40]李徐周,于飞.有效的指纹细化算法[J].计算机工程与设计,2006,4(27),626-628,647
[41]徐建强,李小平.一种改进的细化算法在车牌识别系统中的应用[J].电脑开发与应用,2005,18(7),28-30
[42]杨淑莹,VC++图像处理程序设计(第2版)[M].清华大学出版社,北京交通大学出版社.2005
[43]Rosenfeld. A, Connectivity in digital picture [J], Communications of the ACM,1971: 146-160
[44]Fei Xie, Guili Xu, Yuehua Cheng, Yupeng Tian, An Improved Thinning Algorithm for Human Body Recognition [J]. IEEE Conference Publishing,2009,11(12)
[45]B.K.Jang and R. T. Chin, One-pass Parallel Thinning Analysis, Properties, and Quantitative Evaluation [J], IEEE Trans, Patt. Anal. Machine Intell,1992,14(12): 869-885
[46]R.T.Chin, Hong-Khoon Wan, D.L.Stover, and R.D.Iverson, A one-pass thinning algorithm and its parallel implementation [J], Computer Vision, Graphics, and Image Processing,1987(40):30-40
[47]梅园,孙怀江,夏德深.一种基于改进后模板的图像快速细化算法[J].中国图象图形报,2006,11(9)：1306-1311
[48]王家隆,郭成安.一种改进图像模板细化算法[J].中国图像图形学报,2004,9(3)：297-301
[49]T. Pavlidis, A thinning algorithm for discrete binary images [J], Comput. Graphics Image Processing.1980 (13):142-157
[50]王业琳,宁新宝,尹义龙.指纹图像细化算法的研究[J].南京大学学报(自然科学),2003,39(4)：468-475
[51]U. Montanari, Continuous skeletons from digital images, [J], Assoc. Comput. Mach. 1969(16):534-549
[52]帅金晓,颜永红,彭琰,罗江平.双线性插值图像放大算法优化及硬件实现[J].核电子学与探测技术,2009,29(1)：55-58
[2]啜二勇,国外路面自动检测系统发展综述[J].交通标准化,2009,9(204)：96-99
[3]胡霞光,国内外路面快速检测技术的现状与发展[J].中外公路,2003.23(6)：95-99
[4]Chuo, Er-yong. Development Summary of International Pavement Surface Distress Automatic survey System [J]. Transport Standardization,2009(204):96-99
[5]S Rasmussen, J A Krarup. Non-contact Deflection Measurement at High Speed [C] Lisbon, Portugal:The 6th International Conference on the Bearing Capacity of Roads, Railways and Airfields,2002
[6]Kaseko M S, Lo Z P, Ritchie S G. Comparison of Traditional and Neural Classifiers for pavement crack detection. Journal of Transportation Engineering,1993,119(3):402-418
[7]C.F Chen, Chun Jung Hsu. Development of An Automated Airport Pavement Image Collection System [J]. Journal of Marine Science and Technology,2002,10(1):1-7
[8]啜二勇,国内路面自动检测系统研究历程及展望[J].中国高新技术企业杂志,2009(19)：195-196
[9]张围梁,叶中辰.路面无损检测技术的现状与发展[J].吉林交通科技,2007(3)：60-62
[10]李强.高速公路路面车辙检测、评价与预测技术研究[D].东南大学,2007
[11]Koutsopoulos H N, Downey A B. Primitive-based classification of pavement cracking images [J]. Journal of Transportation Engineering,1993,119(3):402-418
[12]田恩杰,高等级公路路面病害自动检测方法研究[D].吉林大学,2007
[13]申爱琴,庄传仪.山区高速公路沥青路面车辙成因与防治[J].筑路机械与施工机械化,2007(6)：1-4
[14]刘钰,沥青路面车辙深度检测系统研究[D].武汉理工大学,2007
[15]汪恩军,公路车辙自动检测系统研制[D].武汉理工大学,2007
[16]王鑫,唐振民.一种新的自动路面车辙检测方法[J].计算机工程与应用,2008,44(10)：246-248
[17]姚运仕,赵万芹,宋宏勋.沥青路面车辙检测技术的探讨[J].筑路机械与施工机械化,2007(6)：8-10
[18]Mallela R, Wang H. Harmonisiong Automated Rut Depth Measurements:Stage 2, Research Report 277 [R]. Land Transport New Zealand,2006
[19]温智源,贾圣东.基于点激光与线结构光的车辙仪对比[J].山西建筑,2010,36(3)： 350-351
[20]吴海艳,路面车辙特征提取与技术方法研究[D].哈尔滨工业大学.2007
[21]曹冰莹,激光路面检测技术集成与开发研究[D].长安大学,2006
[22]徐卫东,严凯,王鑫.基于结构光的路面车辙检测系统[J].公路交通科技·技术版,2009(6)：49-50,60
[23]Hu GongZhu, Gorge Stonman.3D surface solution using structured light and constraint propagation. IEEE-PAMI,1995,11(4):225-231
[24]任明武,高建贞,杨静宇.一种快速实用的灰度校正算法[J].中国图象图形学报,2002,7(6)：548-552
[25]James D F, Robert A S. Spatially variant contrast enhancement using local ranger modification [J]. SPIE,1983,22(3):378-381
[26]王明辉,沥青路面检测中的图像处理技术研究[D].武汉理工大学,2006.5
[27]周敬,图像分割中阈值法的研究[J].机电技术,2010(1)：39-41,65
[28]高建贞,任明武,唐振民.路面裂缝的自动检测与识别[J].计算机工程,2003,29(2)：49-150
[29]Sun T, Neuvo Y. Detail-preserving median based filters in image processing [J].Pattern Recognition Letters,1994,15(4):341-347.
[30]Wang Z, Zhang D. Progressive switching median filter for the removal of impulse noise from highly corrupted images[J].IEEE Trans on Circuits and Systems Ⅱ:Analog and Digital Signal Processing,1999,46(l):78-80.
[31]XING Cang-ju, WANG Shou-jue, DENG Hao-jiang. A new filtering algorithm based on extremum and median value[J]. Journal of image and Graphics,2001,6(6):533-536.
[32]宋琼琼,贾振红.基于人眼视觉特性的自适应中值滤波算法[J].光电子··激光2009,19(1)：128-130
[33]Pok G, Liu J C, Nair A S, Selective removal of impulse noise based on homogeneity level information[J]. IEEE Trans on Image Processing,2003,12(1):85-92
[34]XU Xiao-in, Miller E 1, CHEN Dong-bin, et al. Adaptive two-pass rank order filer to remove impulse noise in highly corrupted images [J]. IEEE Signal Process Letters,2006, 13(7):413-416
[35]WANG Hong-mei, Li Yan-jun, ZHANG Ke. An improved image filtering method for salt and pepper noise [J]. Journal of Optoelectronics · Laser,2007,18(1):113-116.
[36]赵磊,陈琼,陈中.一种新的改进OPTA细化算法[J].计算机应用,2008,10(10),2638.2642
[37]H. Tamura, A comparison of line thinning algorithms from digital geometry viewpoint [J], in Int. Joint conf. Recognition,1978:715-719
[38]A. Datta and S.K.Parui, A Robust Parrallel Thinning Algorithm for Binary Image [J], Pattern Recognition,1994,27(9):1181-1192
[39]冯星奎,李林艳,颜祖泉.一种新的指纹图像细化算法[J].中国图像图形学报,1999,4(10)：835-838
[40]李徐周,于飞.有效的指纹细化算法[J].计算机工程与设计,2006,4(27),626-628,647
[41]徐建强,李小平.一种改进的细化算法在车牌识别系统中的应用[J].电脑开发与应用,2005,18(7),28-30
[42]杨淑莹,VC++图像处理程序设计(第2版)[M].清华大学出版社,北京交通大学出版社.2005
[43]Rosenfeld. A, Connectivity in digital picture [J], Communications of the ACM,1971: 146-160
[44]Fei Xie, Guili Xu, Yuehua Cheng, Yupeng Tian, An Improved Thinning Algorithm for Human Body Recognition [J]. IEEE Conference Publishing,2009,11(12)
[45]B.K.Jang and R. T. Chin, One-pass Parallel Thinning Analysis, Properties, and Quantitative Evaluation [J], IEEE Trans, Patt. Anal. Machine Intell,1992,14(12): 869-885
[46]R.T.Chin, Hong-Khoon Wan, D.L.Stover, and R.D.Iverson, A one-pass thinning algorithm and its parallel implementation [J], Computer Vision, Graphics, and Image Processing,1987(40):30-40
[47]梅园,孙怀江,夏德深.一种基于改进后模板的图像快速细化算法[J].中国图象图形报,2006,11(9)：1306-1311
[48]王家隆,郭成安.一种改进图像模板细化算法[J].中国图像图形学报,2004,9(3)：297-301
[49]T. Pavlidis, A thinning algorithm for discrete binary images [J], Comput. Graphics Image Processing.1980 (13):142-157
[50]王业琳,宁新宝,尹义龙.指纹图像细化算法的研究[J].南京大学学报(自然科学),2003,39(4)：468-475
[51]U. Montanari, Continuous skeletons from digital images, [J], Assoc. Comput. Mach. 1969(16):534-549
[52]帅金晓,颜永红,彭琰,罗江平.双线性插值图像放大算法优化及硬件实现[J].核电子学与探测技术,2009,29(1)：55-58