基于多尺度图像分析的路面病害检测方法研究与分析
|
摘要
近年来,基于CCD摄像机的路面病害自动检测系统在公路养护事业中获得了广泛的应用。而病害自动检测这一模块的检测效果将直接影响到整个系统的检测精度,虚检和漏检都会影响整个道路状况的评估,因此设计准确有效的自动检测算法是至关重要的一个环节。传统的大部分自动检测算法都是在固定尺度上对路面图像进行分析,这类方法不能表现出图像本身固有的多尺度特性,也就限制了分析结果的准确性,而本文采用一种模拟人类视觉感知外界事物过程的多尺度图像分析方法对路面图像进行分析,围绕着这个核心从路面图像的去噪、增强和裂缝目标的提取三个方面进行研究。
针对路面图像表面纹理丰富、背景噪声多的特性,本文先从数学形态学尺度空间出发,构造了一种多尺度形态学滤波器,根据路面图像的特性选择了合适的形态运算类型、结构元素以及恰当的尺度,并根据不同尺度抗噪性能的不同采用了合适的权值分配方案,该方法能很好地在滤除背景噪声的同时保持目标边缘。
本文又从非线性尺度空间出发,针对P-M正则化模型中的高斯预平滑会带来图像边缘位置漂移的缺点,对模型扩散方程的扩散系数进行修改,将形态学算子引入扩散方程中,提出了一种基于形态学算子的P-M模型,模型中梯度阈值K采用一种鲁棒统计的方法自适应地给出。新的扩散模型在有效去除噪声的同时具有更好的目标边缘保持能力,并且能够抑制边界的移动。
针对路面图像中裂缝信息弱小,与背景对比度低的情况,仅仅滤波处理是不够的,鉴于此,本文将基于非线性尺度空间的图像增强技术应用到路面图像中,并针对裂缝的线形纹理结构,采用一种能够反应图像纹理方向的结构张量与Osher-Rudin冲激滤波模型结合的相干增强冲激滤波模型,该模型不但能够有效地锐化目标边缘,增强图像中目标和背景的对比度,更重要的是增强了裂缝的线形纹理结构。
在裂缝信息的提取上,本文针对背景光照不均的图像设计了一种基于多尺度形态学梯度的裂缝目标检测方法,通过检测图像的局部突变信息来避免因整体灰度不均而导致的错分割现象。该方法综合了大小尺度的梯度信息,克服了常规算法易受噪声、杂物以及光照条件等干扰的问题,对不同路况路面中的裂缝均能取得良好的检测结果。
Nowadays, the automatic pavement distress detection system based on CCD camera has been developed greatly in the road maintenance project. And the automatic distress detection module will directly affect the measurement accuracy of the whole system, wrong detected and missed will both affect the assessment of the entire road conditions, so designing accurate and effective automatic detection algorithm is a crucial link. Most of traditional auto-detection algorithms analyze the road image in a fixed scale, such methods can not show inherent multi-scale characteristics of images, so the accuracy of analysis results is limited. In this paper, the multi-scale image analysis method which simulates human visual perception of outside things is used to analyze pavement surface images, and around the core we study three areas: pavement surface image denoising, enhancement and crack targets extracting.
To aim at the characteristics of pavement surface's rich texture and much background noise, this paper firstly starts from the mathematical morphology scale space, constructs a multi-scale morphological filter, and selects the appropriate morphological computing type, structure element as well as appropriate scale according to the characteristics of road images. And appropriate weight distribution is used according to different anti-noise performance in different scales. This method can be very good at filtering out background noise while maintaining the edge of the target.
Then this paper starts from the nonlinear scale space, to aim at the shortcomings of the Gaussian pre-smoothing in the P-M regularization model will bring about the edge position drifting, we modify the diffusion coefficient of the diffusion equation, bring morphological operators to the diffusion equation, then a new P-M model based on morphological operators is formed, the gradient threshold K in the model is given adaptively by a robust statistics method. New diffusion model can remove noise effectively while have better edge retention capacity, and can inhibit the drifting of border.
To the situation of small crack information with low contrast to background, filtering is not enough, so in this paper image enhancement technology based on nonlinear scale space is applied to road surface images, and to aim at the linear texture structure of cracks, this paper adopts a coherence-enhancing shock filter which combines structure tensor with the Osher-Rudin shock filter model. This model is not only able to effectively sharpen target's edge but also enhance the contrast of the target and background and, more importantly, enhance the linear texture structure of the cracks.
About the extraction of crack information, this paper design a crack detecting method based on multi-scale morphological gradient to aim at the images with uneven background illumination, it avoids wrong segmentation phenomenon caused by uneven overall gray by detecting the local mutations information. This method combines large scale gradient information with small scale's, it overcomes the problem of conventional method's susceptible to noise, debris and lighting conditions, and has achieved good detecting results of cracks in different road conditions.
针对路面图像表面纹理丰富、背景噪声多的特性,本文先从数学形态学尺度空间出发,构造了一种多尺度形态学滤波器,根据路面图像的特性选择了合适的形态运算类型、结构元素以及恰当的尺度,并根据不同尺度抗噪性能的不同采用了合适的权值分配方案,该方法能很好地在滤除背景噪声的同时保持目标边缘。
本文又从非线性尺度空间出发,针对P-M正则化模型中的高斯预平滑会带来图像边缘位置漂移的缺点,对模型扩散方程的扩散系数进行修改,将形态学算子引入扩散方程中,提出了一种基于形态学算子的P-M模型,模型中梯度阈值K采用一种鲁棒统计的方法自适应地给出。新的扩散模型在有效去除噪声的同时具有更好的目标边缘保持能力,并且能够抑制边界的移动。
针对路面图像中裂缝信息弱小,与背景对比度低的情况,仅仅滤波处理是不够的,鉴于此,本文将基于非线性尺度空间的图像增强技术应用到路面图像中,并针对裂缝的线形纹理结构,采用一种能够反应图像纹理方向的结构张量与Osher-Rudin冲激滤波模型结合的相干增强冲激滤波模型,该模型不但能够有效地锐化目标边缘,增强图像中目标和背景的对比度,更重要的是增强了裂缝的线形纹理结构。
在裂缝信息的提取上,本文针对背景光照不均的图像设计了一种基于多尺度形态学梯度的裂缝目标检测方法,通过检测图像的局部突变信息来避免因整体灰度不均而导致的错分割现象。该方法综合了大小尺度的梯度信息,克服了常规算法易受噪声、杂物以及光照条件等干扰的问题,对不同路况路面中的裂缝均能取得良好的检测结果。
Nowadays, the automatic pavement distress detection system based on CCD camera has been developed greatly in the road maintenance project. And the automatic distress detection module will directly affect the measurement accuracy of the whole system, wrong detected and missed will both affect the assessment of the entire road conditions, so designing accurate and effective automatic detection algorithm is a crucial link. Most of traditional auto-detection algorithms analyze the road image in a fixed scale, such methods can not show inherent multi-scale characteristics of images, so the accuracy of analysis results is limited. In this paper, the multi-scale image analysis method which simulates human visual perception of outside things is used to analyze pavement surface images, and around the core we study three areas: pavement surface image denoising, enhancement and crack targets extracting.
To aim at the characteristics of pavement surface's rich texture and much background noise, this paper firstly starts from the mathematical morphology scale space, constructs a multi-scale morphological filter, and selects the appropriate morphological computing type, structure element as well as appropriate scale according to the characteristics of road images. And appropriate weight distribution is used according to different anti-noise performance in different scales. This method can be very good at filtering out background noise while maintaining the edge of the target.
Then this paper starts from the nonlinear scale space, to aim at the shortcomings of the Gaussian pre-smoothing in the P-M regularization model will bring about the edge position drifting, we modify the diffusion coefficient of the diffusion equation, bring morphological operators to the diffusion equation, then a new P-M model based on morphological operators is formed, the gradient threshold K in the model is given adaptively by a robust statistics method. New diffusion model can remove noise effectively while have better edge retention capacity, and can inhibit the drifting of border.
To the situation of small crack information with low contrast to background, filtering is not enough, so in this paper image enhancement technology based on nonlinear scale space is applied to road surface images, and to aim at the linear texture structure of cracks, this paper adopts a coherence-enhancing shock filter which combines structure tensor with the Osher-Rudin shock filter model. This model is not only able to effectively sharpen target's edge but also enhance the contrast of the target and background and, more importantly, enhance the linear texture structure of the cracks.
About the extraction of crack information, this paper design a crack detecting method based on multi-scale morphological gradient to aim at the images with uneven background illumination, it avoids wrong segmentation phenomenon caused by uneven overall gray by detecting the local mutations information. This method combines large scale gradient information with small scale's, it overcomes the problem of conventional method's susceptible to noise, debris and lighting conditions, and has achieved good detecting results of cracks in different road conditions.
引文
[1]Velisky S A.Kirschke K R.Design considerations for automated pavement crack selling machinery[R].Proc.2nd Int.Conf.on Applications of Advanced Technologies in Transp.Engrg.1994.
[2]Li L ,Chan P, Rao A.Flexible pavement distress evaluation using image analysis[J].1993(119):402-418.
[3]Grivas D A, Bhagvati C, Skolnick M M.Feasibility of automating pavement distress assessment using mathematical morphology [J].Transportation Research Record.1994(1435): 52-58.
[4]Cheng H D, Rong Jim.Novel approach to pavement cracking detection based on fuzzy set theory[J].ASCE.1999(13):270-280.
[5]D.Meignen, M.Bernadet, H.Briand.One Application of Neural Networks for Detection of Defects Using Video Data Bases: Identification of Road Distress[R].IEEE.1997.
[6]Bhagvati.C, Skonlnick.M.M, and Grivsa.D.A.Gaussian normalization of morphological size distributions for increasing sensitivity to texture variations and its application to pavement distress classification[R].IEEE Computer Vision and Pattern Recognition Conference.1994.
[7]S.Paquis, V.Legeay, H.Konik, J.Charrier.Multiresolution Texture Analysis applied to Road Surface Inspection[J]SPIE.1999(3652): 242-249.
[8]F.P.Lovergine, A.Branca, GAttolico, A.Distante.Leather Inspection by Oriented Texture Analysis with a Morphological Approach[R].IEEE.1997.
[9]Cheng Hengda.Automated real-time pavement distress detection using fuzzy logic and neural network[J].SPIE Proceeding 1994(2946): 140-151.
[10]H.D.Cheng, Jim-Rong Chen.A novel fuzzy logic approach to pavement distress detection[J].SPIE.1994(2946): 97-108.
[11]Naoki Tanaka, Kenji Uematsu.A crack detection method in road surface images using morphology[R].MVA'98 LAPR Workshop on Machine Vision.1998.
[12]A.Witkin.Scale-space Filtering[R].In Int.Joint Conf.on AI, Karlsruhe.1983: 1019-1022.
[13]W.Liu, T.F.Fwa, and Z.Zhao.Wavelet analysis for pavement roughness studies[R].Proc.8th Int.Conf.on Application of Advanced Technologies in Transportation Engineering.2004: 455-459.
[14]W.Liu,T.F.Fwa,and Z.Zhao.Wavelet analysis and interpretation of road roughness[J].J.Transp.Eng.2005,131(2):120-130.
[15]N.Oguri,K.Himeno,A.Kawamura,and R.Nakamura.Application of wavelet analysis on evaluation of roughness of pavement surfaces[R].Proc.4th Int.Symp.on Pavement Surface Characteristics of Roads and Airfields.2000:57-66.
[16]刘曙,罗予频,杨士元.基于多尺度形态学的红外图像边缘检测方法[J].计算机应用.2007,27(4):971-975.
[17]Tony Lindeberg.Bart M Ter Harr Romeny.Linear scale space[M].Netherlands:Kluwer Academic Publishers.1994.
[18]Marr D C,Hildreth E C.Theory of edge detection[A].Proc.R.Soc.London.1980,207(B):187-217.
[19]张亶,陈刚.基于偏微分方程的图像处理[M].北京:高等教育出版社.2004.
[20]Koenderink J J.The structure of image[J].Biological Cybernetics,1984,50:363-370.
[21]Perona P,Malik J.Scale-space and edge detection using anisotropic diffusion[J].IEEE Trans.Pattern Analysis and Machine Intelligence,1999,12(7):1163-1173.
[22]Alvarez L,et al.Axioms and Fundamental Equations of Image Processing[J].Arch.Rational Mech.Anal.1993,123:199-257.
[23]Park K R,Lee C N.Scale-space using mathematical morphology[J].IEEE Trans.Pattern Analysis and Machine Intelligence,1996,18(11):1121-1126
[24]Canny J.A computational approach to edge detection[J].IEEE Trans.Pattern Analysis and Machine Intelligence,1986,8(6):679-698.
[25]Guichard F,Morel J M.Image iterative smoothing and P.D.E's[R].Lecture Notes,Beijing,1999.
[26]Guillermo S.Geometric Partial Differential Equations and Image Analysis[M].Cambrige:Cambrige University Press,2001.
[27]You Yu-Li et al.Behavioral Anisotropic Diffusion in Image Processing[J].IEEE Transaction on Image Processing,1996:5(11):1539-1553.
[28]Matheron G.Random Sets and Integral Geometry[M].New York:Wiley 1975.
[29]Serra.Image Analysis and Mathematical Morphology[M].Academic Press.1982.
[30]Lax P D.Numerical solution of partial differential equations[J].Am.Math.Mon.1965:72(2):74-84.
[31]Goutsias J.,Vincent L.,and Bloomberg D.S.,Image mathematical morphology and its applications to image and signal processing[M].Boston:Kluwer Academic Publishers,2000.
[32]J Serra.Morphological filtering:An overview[J].Signal processing,1994,38(1):3-11.
[33]S Mukhopadhyay,B Chanda.An edge preserving noise smoothing technique using multiscale morphology[J].Signal Processing,2002,82(4):527-544.
[34]F.Catte,P.L.Lions,J.M.Morel,T.Coll.Image selective smoothing and edge detection by nonlinear diffusion[J].SIAM Journal of Numerical Analysis.1992,29(1):182-193.
[35]C.Andrew Segall and Scott T.Acton.Morphological Anisotropic Diffusion[R].IEEE.Image Processing,1997(3):348-351.
[36]Guillermo Sapiro.Geometric Partial Differential Equations and Image Analysis[M].Cambridge:2003.
[37]D.A.F.Florencio and R.W.Schafer.Homotopy and Critical Morphological Sampling[R].Proc of SPIE Symposium on Visual Communications and Image Processing,Chicago,September 1994.
[38]W.K.Pratt.Digital Image Processing[M].New York:Wiley,pp.495-501,1978.
[39]Weickert J.Anisotropic Diffusion in Image Processing[M].Stuttgart.Germany:Teubner-Verlag.1998.
[40]Weickert J.Coherence Enhancing Diffusion Filtering[J].International journal of Computer Vision,1999,31(2/3):111-127.
[41]S.J.Osher and L.I.Rudin.Feature-oriented image enhancement using shock filters[J].SIAM J.Numer.Anal.1990(27):919-940.
[42]L.Remaki,M.Cheriet.Enhanced and restored signals as a generalized solution for shock filter models,Part Ⅰ-existence and uniqueness result of the Cauchy problem[J].Journal of Mathematical Analysis and Applications,2003,279:189-209.
[43]G.Gilboa,N.Sochen,Y.Y.Zeevi.Regularized shock filters and complex diffusion[J].Springer-Verlag,2002(2350):399-413.
[44]L.Alvarez,L.Mazorra.Signal and image restoration using shock filters and anisotropic diffusion[J].SIAM Journal of Numerical Analysis.1994,31(2):590-605.
[45]K.W.Morton,D.F.Mayers.偏微分方程数值解[M].北京:人民邮电出版社,2006.
[46]J.Weickert.Coherence-enhancing shock filters.Lecture Notes in Computer Science,2781,2003,Springer,Berlin.
[47]S.J.Osher,J.A.Sethian.Fronts propagation with curvature dependent speed:algorithms based on Hamilton Jacobi formulations,J.Computational Phys.79,12-49,1988.
[48]陈虎,王守尊,周朝辉.基于数学形态学的图像边缘检测方法研究[J].2004(2):112-114.
[49]章毓晋.图像分析[M].北京:清华大学出版社,2003.
[50]P.Soille.形态学图像分析原理与应用[M].北京:清华大学出版社,2008.
[2]Li L ,Chan P, Rao A.Flexible pavement distress evaluation using image analysis[J].1993(119):402-418.
[3]Grivas D A, Bhagvati C, Skolnick M M.Feasibility of automating pavement distress assessment using mathematical morphology [J].Transportation Research Record.1994(1435): 52-58.
[4]Cheng H D, Rong Jim.Novel approach to pavement cracking detection based on fuzzy set theory[J].ASCE.1999(13):270-280.
[5]D.Meignen, M.Bernadet, H.Briand.One Application of Neural Networks for Detection of Defects Using Video Data Bases: Identification of Road Distress[R].IEEE.1997.
[6]Bhagvati.C, Skonlnick.M.M, and Grivsa.D.A.Gaussian normalization of morphological size distributions for increasing sensitivity to texture variations and its application to pavement distress classification[R].IEEE Computer Vision and Pattern Recognition Conference.1994.
[7]S.Paquis, V.Legeay, H.Konik, J.Charrier.Multiresolution Texture Analysis applied to Road Surface Inspection[J]SPIE.1999(3652): 242-249.
[8]F.P.Lovergine, A.Branca, GAttolico, A.Distante.Leather Inspection by Oriented Texture Analysis with a Morphological Approach[R].IEEE.1997.
[9]Cheng Hengda.Automated real-time pavement distress detection using fuzzy logic and neural network[J].SPIE Proceeding 1994(2946): 140-151.
[10]H.D.Cheng, Jim-Rong Chen.A novel fuzzy logic approach to pavement distress detection[J].SPIE.1994(2946): 97-108.
[11]Naoki Tanaka, Kenji Uematsu.A crack detection method in road surface images using morphology[R].MVA'98 LAPR Workshop on Machine Vision.1998.
[12]A.Witkin.Scale-space Filtering[R].In Int.Joint Conf.on AI, Karlsruhe.1983: 1019-1022.
[13]W.Liu, T.F.Fwa, and Z.Zhao.Wavelet analysis for pavement roughness studies[R].Proc.8th Int.Conf.on Application of Advanced Technologies in Transportation Engineering.2004: 455-459.
[14]W.Liu,T.F.Fwa,and Z.Zhao.Wavelet analysis and interpretation of road roughness[J].J.Transp.Eng.2005,131(2):120-130.
[15]N.Oguri,K.Himeno,A.Kawamura,and R.Nakamura.Application of wavelet analysis on evaluation of roughness of pavement surfaces[R].Proc.4th Int.Symp.on Pavement Surface Characteristics of Roads and Airfields.2000:57-66.
[16]刘曙,罗予频,杨士元.基于多尺度形态学的红外图像边缘检测方法[J].计算机应用.2007,27(4):971-975.
[17]Tony Lindeberg.Bart M Ter Harr Romeny.Linear scale space[M].Netherlands:Kluwer Academic Publishers.1994.
[18]Marr D C,Hildreth E C.Theory of edge detection[A].Proc.R.Soc.London.1980,207(B):187-217.
[19]张亶,陈刚.基于偏微分方程的图像处理[M].北京:高等教育出版社.2004.
[20]Koenderink J J.The structure of image[J].Biological Cybernetics,1984,50:363-370.
[21]Perona P,Malik J.Scale-space and edge detection using anisotropic diffusion[J].IEEE Trans.Pattern Analysis and Machine Intelligence,1999,12(7):1163-1173.
[22]Alvarez L,et al.Axioms and Fundamental Equations of Image Processing[J].Arch.Rational Mech.Anal.1993,123:199-257.
[23]Park K R,Lee C N.Scale-space using mathematical morphology[J].IEEE Trans.Pattern Analysis and Machine Intelligence,1996,18(11):1121-1126
[24]Canny J.A computational approach to edge detection[J].IEEE Trans.Pattern Analysis and Machine Intelligence,1986,8(6):679-698.
[25]Guichard F,Morel J M.Image iterative smoothing and P.D.E's[R].Lecture Notes,Beijing,1999.
[26]Guillermo S.Geometric Partial Differential Equations and Image Analysis[M].Cambrige:Cambrige University Press,2001.
[27]You Yu-Li et al.Behavioral Anisotropic Diffusion in Image Processing[J].IEEE Transaction on Image Processing,1996:5(11):1539-1553.
[28]Matheron G.Random Sets and Integral Geometry[M].New York:Wiley 1975.
[29]Serra.Image Analysis and Mathematical Morphology[M].Academic Press.1982.
[30]Lax P D.Numerical solution of partial differential equations[J].Am.Math.Mon.1965:72(2):74-84.
[31]Goutsias J.,Vincent L.,and Bloomberg D.S.,Image mathematical morphology and its applications to image and signal processing[M].Boston:Kluwer Academic Publishers,2000.
[32]J Serra.Morphological filtering:An overview[J].Signal processing,1994,38(1):3-11.
[33]S Mukhopadhyay,B Chanda.An edge preserving noise smoothing technique using multiscale morphology[J].Signal Processing,2002,82(4):527-544.
[34]F.Catte,P.L.Lions,J.M.Morel,T.Coll.Image selective smoothing and edge detection by nonlinear diffusion[J].SIAM Journal of Numerical Analysis.1992,29(1):182-193.
[35]C.Andrew Segall and Scott T.Acton.Morphological Anisotropic Diffusion[R].IEEE.Image Processing,1997(3):348-351.
[36]Guillermo Sapiro.Geometric Partial Differential Equations and Image Analysis[M].Cambridge:2003.
[37]D.A.F.Florencio and R.W.Schafer.Homotopy and Critical Morphological Sampling[R].Proc of SPIE Symposium on Visual Communications and Image Processing,Chicago,September 1994.
[38]W.K.Pratt.Digital Image Processing[M].New York:Wiley,pp.495-501,1978.
[39]Weickert J.Anisotropic Diffusion in Image Processing[M].Stuttgart.Germany:Teubner-Verlag.1998.
[40]Weickert J.Coherence Enhancing Diffusion Filtering[J].International journal of Computer Vision,1999,31(2/3):111-127.
[41]S.J.Osher and L.I.Rudin.Feature-oriented image enhancement using shock filters[J].SIAM J.Numer.Anal.1990(27):919-940.
[42]L.Remaki,M.Cheriet.Enhanced and restored signals as a generalized solution for shock filter models,Part Ⅰ-existence and uniqueness result of the Cauchy problem[J].Journal of Mathematical Analysis and Applications,2003,279:189-209.
[43]G.Gilboa,N.Sochen,Y.Y.Zeevi.Regularized shock filters and complex diffusion[J].Springer-Verlag,2002(2350):399-413.
[44]L.Alvarez,L.Mazorra.Signal and image restoration using shock filters and anisotropic diffusion[J].SIAM Journal of Numerical Analysis.1994,31(2):590-605.
[45]K.W.Morton,D.F.Mayers.偏微分方程数值解[M].北京:人民邮电出版社,2006.
[46]J.Weickert.Coherence-enhancing shock filters.Lecture Notes in Computer Science,2781,2003,Springer,Berlin.
[47]S.J.Osher,J.A.Sethian.Fronts propagation with curvature dependent speed:algorithms based on Hamilton Jacobi formulations,J.Computational Phys.79,12-49,1988.
[48]陈虎,王守尊,周朝辉.基于数学形态学的图像边缘检测方法研究[J].2004(2):112-114.
[49]章毓晋.图像分析[M].北京:清华大学出版社,2003.
[50]P.Soille.形态学图像分析原理与应用[M].北京:清华大学出版社,2008.