【作者】 田恩杰；

【导师】 王国强；

【作者基本信息】 吉林大学 ， 机械设计及理论， 2007， 硕士

【摘要】 本文结合吉林省科技厅高新技术项目“路基路面智能集成检测车”,研究了基于图像处理的沥青公路路面裂缝病害检测方法,并开发了检测软件。本文综述了公路路面病害检测技术的研究现状,并对沥青公路路面破损及其检测方法进行了介绍。采用平稳小波阈值去噪方法,对实际采集的路面病害图像进行了去噪处理,利用侧抑制原理对去噪后的图像进行了图像增强,图像的质量得到了明显的改善。基于模糊C-均值聚类和最大类间方差法,提出了一种自适应阈值图像分割方法,对路面病害图像进行了处理。设计了BP神经网络识别系统,对分割处理后的病害图像进行识别,实验表明该网络能够较好的对路面裂缝进行识别。本文的研究结果可直接应用于公路路面的破损检测及病害识别,也可为其他领域的图像分析研究提供参考。更多还原

【Abstract】 The road is critical infrastructure to support national economic development and social progress, and it holds the important status in the national economy and the lives of the people. As the rapid developing of road construction and the consummating of road network system, the road maintenance management is becoming increasingly important. The detection of road condition which is an important part of the road maintenance management is facing higher requirements. The traditional manual methods can’t meet the need of pavement’s development now, because they have the following shortcomings such as low efficiency, affecting transportation and non-precision. How to realize accurately, rapidly, non-destructively road condition detecting is the developing direction.Pavement crack which is the main form of the pavement surface distress is the key detection project. Pavement cracks which are in a variety of forms are the key point and difficulty in pavement surface distress detection. At present, many researches on the pavement surface distress detection are focused on it. Depending on the high-tech project of Jilin Provincial Science&Technology Department–“The Integrated Detection Vehicle of subgrade and Pavement”, this thesis studied pavement crack distress detection methods which are based on image processing. In this thesis, stationary wavelet thresholding method on the pavement image denoising is proposed, then, lateral inhibition principle is used for the image enhancing. An adaptive thresholding segmentation method based on the fuzzy c-means clustering and Otsu law is used in the road image segmentation processing. Pavement Cracks are identified by the BP neural network method. The main content of this thesis summarized as follows:(1) It introduces the status and development trends of the technology of pavement surface distress detection on the basis of referring to domestic and foreign literature.(2) It briefly introduces the asphalt road surface damaged knowledge, including the types of road damage, the causes, characteristics and hazards of several types of cracks distress, the evaluation methods of road damaged condition and pavement condition index PCI calculation method etc. In addition, the thesis also introduces the imaging characteristics of the pavement image.(3) Stationary wavelet thresholding method on the pavement image denoising is proposed in the thesis. This method is very effective by comparing with median filter, neighborhood weighted mean filter and DWT thresholding method. After image denoising, lateral inhibition principle is used for the image enhancing. The crack distress information is enhanced.(4) An adaptive thresholding segmentation method based on the fuzzy c means clustering and Otsu law is used in the road image segmentation processing. Experimental results show that this method is effective than widely used Otsu law. The paper uses a method that extracts information in crack seed image which is built according to the direction template of crack type.(5) BP network is simple and practical, widely used in pattern recognition, classification, function approximation, data compression etc.. The paper brings a pavement crack image type classification method based on BP neural network using domestic and foreign studying results for reference. We design a three-layer BP neural network using six features extracting from crack seed image as the input and five type of pavement crack as the output. These five features are transverse crack, longitudinal crack, alligator crack, block crack and no crack. In order to train and evaluate the network, we collected 150 and 100 samples images as the training set and test set respectively. The results show that the network can be used for identificating pavement cracks, and the overall recognition rate over 90%.(6) The integrated detection testing platform of subgrade and pavement is introduced, including components, processes and the composition of some subsystems. In the final, the main reseach of this thesis is summarized, and the subsequent research work is presented.This thesis studies on the high-grade asphalt road surface crack detection. Some of methods used in this thesis are novel in theory and feasible in practice. The author hopes that this paper is helpful to the follow-up research.更多还原