【作者】 孙晓明；

【导师】 刘宛予；

【作者基本信息】 哈尔滨工业大学 ， 仪器科学与技术， 2012， 博士

【摘要】 随着高速公路建设的快速发展，公路检测与养护管理已成为我国公路建设领域的重要任务。裂缝是评价路面质量最重要的参数之一，是大部分病害的早期表现形式，直接影响着公路使用寿命和行车安全，及早发现裂缝并进行维护可以及时避免裂缝进一步发展造成的严重影响。基于结构光的公路路面裂缝三维检测技术具有数据精度高、特征丰富，对路面阴影、黑斑以及对随机噪声不敏感等优势，很好的克服了基于二维灰度图像的路面裂缝检测技术存在的对于路面修补、油污、轮胎痕迹、阴影、光照不均等干扰因素敏感的问题，有效的提高了路面裂缝识别率，成为路面裂缝自动化检测领域新的发展方向。本课题围绕结构光三维裂缝检测技术这一主题，对结构光三维裂缝检测系统中存在的光条中心提取无法兼顾速度与精度的问题、三维路面轮廓信号特征提取问题以及裂缝检测系统的决策可靠性问题这三个关键问题进行研究。裂缝在结构光检测时体现为光条的微小变形信息，光条中心提取方法须具有较高的精度才能保证微小变形信息的精确提取，才能确保三维数据中裂缝及其他病害特征提取的准确性，而面对海量的公路路面数据，提高光条中心提取的速度也是至关重要的；经过光条中心提取、标定之后得到的三维路面轮廓信号病害特征丰富，只有将各个病害从主轮廓中分离，并且分离时不破坏主轮廓信息，才能保证裂缝及平整度等其他参数计算的准确性，而病害与主轮廓的良好分离一直是路面轮廓信号处理的难点；在检测裂缝的过程中，不可避免的会出现裂缝漏检和误判的情况，分析裂缝检测的可靠性是裂缝检测领域重要的研究问题。因此针对目前已有光条中心提取方法无法满足高精度高速度的问题、三维路面轮廓信号特征较难分离的问题、裂缝检测系统决策可靠性问题，本文研究了相应的解决办法。本文主要研究工作如下：（1）结构光光条中心提取的精度和速度直接影响着结构光三维裂缝检测系统中三维数据精度和处理速度。针对目前结构光光条中心提取方法不能兼顾高精度和高速度的问题并结合路面光条图像复杂多变、中心提取困难的特点，提出了基于脊线跟踪与海森(Hessian)矩阵相结合的光条中心提取方法。提出的方法利用Radon变换获取光条感兴趣区域(ROI)，在ROI区域中利用Hessian矩阵计算光条的亚像素坐标，利用脊线跟踪方法提取图像的光条中心线。实验表明，本文方法兼有Hessian矩阵提取光条精度高、抗噪性好和脊线跟踪处理图像点数少，处理速度快的优点。（2）针对目前路面轮廓信号裂缝特征提取方法无法实现裂缝准确检测以及裂缝与主轮廓良好分离的问题，提出了基于稀疏分解的公路路面裂缝特征提取方法。该方法将稀疏分解理论引入到了路面轮廓信号处理领域中，根据路面裂缝信号特征，建立匹配的过完备原子库，利用匹配追踪算法实现裂缝特征参数的准确检测以及路面裂缝特征与路面主轮廓的分离。实验结果表明，本文提出的方法能够在不破坏路面主轮廓完整性的基础上，实现路面裂缝特征信息与主轮廓的良好分离，实现路面裂缝特征参数的准确计算。（3）针对结构光三维裂缝检测系统决策可靠性问题，建立了基于假设检验的公路路面裂缝识别决策模型。首先综合分析了基于结构光的公路路面裂缝三维检测技术中影响决策可靠性的因素，提出了一种基于假设检验的公路路面裂缝识别决策模型。该模型确定了各个因素之间、各个因素与检测正确率之间的关系，能够为实际路面裂缝检测起到指导作用，仿真实验与实际路面实验验证了该模型的准确性与有效性。更多还原

【Abstract】 With the rapid development of highway construction, pavement inspection andmaintenance management have become an important task for China highwayconstruction. Cracks are one of the most important parameters to evaluate thesurface quality and the early manifestations of most disease, which will affect thelifespan of the pavement and traffic safety directly. It can avoid serious problemscaused by the crack further development if we detect cracks early and maintainthem in time. The major issue with pure2D detection technique is that they can notdiscriminate dark areas not caused by pavement distress such as recent fillings, oilspills, tire marks, shadows and uneven poor illumination. Three-dimensionalpavement crack detection technique based on structured light has the advantages ofhigh accuracy, rich feature and is robust to the shadows, black spots and randomnoise, which can effectively improve the recognition rate of pavement surfacecracks. So this method becomes the new development direction in the field ofpavement crack automatic inspection.The main purpose of this project is to study key issues in the processingtechniques of pavement surface crack data based on structured light and focus onthree key technologies of3D crack detection system based on the structured light,including unable to balance speed and accuracy of three-dimensional data detection,pavement profile signal feature extraction and decision reliabillity problem of crackdetection system. Cracks are reflected in the small deformation of the light stripeduring structured light detection, the extraction of light stripe center must havehigh-precision resistance in order to ensure accurate extraction of the smalldeformation information, in the same time in order to ensure the accuracy of thecracks and other disease characteristics in the three-dimensional data extracted.While the surface data of the pavement is massive, to speed up light stripe centerextraction is also essential. After the light stripe extraction and calibration, thethree-dimensional pavement profile signals have rich disease characteristics. Onlyeach disease has been separated from the main profile well and the information ofmain profile has not been damaged, we can ensure the accuracy of the calculation ofother parameters such as cracks and IRI. Good separation of the disease and themain profile has been the difficulty in the area of profile signal processing. In thedetection process of cracks, inevitably there will be cracks undetected and amiscarriage of justice, analysis of the reliability of crack detection is an importantresearch in crack detection area. This study focus on the problem of Light stripecenter extraction method can not meet the high-precision and high-speed signalcharacteristics, three-dimensional pavement profile more difficult separation problems, the reliability of crack detection system decision-making.The main research works in this paper are as follows:Accuracy and speed of structured light center extraction impact on the dataprocessing accuracy and speed of three-dimensional detection system. Consideringthat structured light center extraction method can not meet high precision, highspeed and pavement image is complex and difficult to extract laser stripe center. Wepropose the laser stripe center extraction method based on ridge-tracing withHessian matrix. This method possess the Hessian’s merits of higher extractionaccuracy, better performance of resistance to noise and less image points usingridge-tracing. Utilize Radon transform to get ROI of laser stripe and trace ridge lineof laser stripe in ROI. In ROI area, calculate the Hessian matrix in a certainneighborhood along the normal direction to find some points satisfying the ridgepoint conditions as the center point of laser stripe. The experiment demonstratesthat our method has better performance in precision, speed and resistance of thenoise.For the problems of cracks are difficult to extract accurately and to beseparated from main profile, we propose the feature extraction method of thepavement cracks, so as to achieve crack precise detection and the separation frommain profile. First, it introduces sparse decomposition theory applied to the field ofpavement profile signal processing and builds over-complete atoms dictionary inaccordance with characteristics of pavement cracks signal.Then signal is separatedby learning in this mixed dictionary with a matching pursuit (MP) algorithm.Experiments show that this method can separate crack characterizes informationand accurately detect the distress parameters without losing the main outline.For crack decision-making reliability problems of structured light3D crackdetection system, a decision model for the laser scanning pavement crack detectionsystem based on the hypothesis test is proposed. First the factors contributed tothese errors in laser scanning system are firstly analyzed, and then a decision modelfor the laser scanning pavement crack detection system based on the hypothesis testis proposed. This model build the relationship between the contribution factors andcrack detection accuracy and can provide guidance on the pavement crack detectionand has practical value. Simulation and the actual road experiment verify theaccuracy and validity of the model.更多还原