基于Legendre矩和分数阶积分的复杂路面裂缝检测及算法评价

发布时间：2019-01-26 17:26

【摘要】：随着公路交通的快速发展,路面状况检测与养护已成为我国公路建设的首要任务。而裂缝是衡量路面质量的重要指标之一,因此利用数字图像处理技术进行路面裂缝检测已经成为该领域研究的热点。在实际检测过程中,由于路面情况复杂,使得采集的路面图像中存在诸如油污、阴影、光照不均和随机噪声等干扰因素。在这类情形下,现有的裂缝检测方法存在误判和漏检问题,不能很好地满足检测需求,无法获得较准确和全面的裂缝信息,进而不能及时对路面进行有效的养护管理。针对上述问题,本文主要从以下四个方面对复杂路面裂缝检测进行深入研究:路面图像增强、裂缝区域提取、断点连接及参数计算、裂缝检测算法评价。(1)考虑到复杂路面裂缝图像具有干扰噪声多、阴影和光照不均等特征,本文提出了基于小波分析的路面裂缝图像增强算法。采用非线性变换对低频分量进行增强处理,同时对高频分量进行小波阈值去噪处理,抑制高频部分的噪声信息,然后通过小波重构得到增强后的图像。实验结果表明该算法不仅很好地增强了路面图像的对比度,而且在抑制噪声的同时能够最大限度地保留裂缝的边缘细节。(2)针对背景模糊、光照不均及阴影等复杂路面裂缝图像,本文提出了一种基于Legendre矩和分数阶积分的路面裂缝提取方法,首先通过Legendre矩找到最佳的类似于参考图像的连通域,即找到最优的分数阶积分阶次;然后利用一个最优阶次的分数阶积分掩膜来处理图像,使图像中像素的灰度级减少;最后计算分数阶积分算子处理后图像的直方图,根据直方图确定最优阈值以便提取裂缝信息。该方法充分地利用了分数阶积分的性质,考虑了像素的空间分布,增加了图像的均匀性,不但去除了大量的噪声干扰点,而且较完整地提取出裂缝区域。(3)对于所提取的裂缝不连续、断裂等现象,本文采用一种基于区域搜寻的裂缝连接算法进行裂缝断点连接,先依据深度优先原则进行裂缝邻域搜索,再根据连通性原理去除断点。并对连接后的裂缝进行长度和宽度的测量,分别采用骨架提取法和二阶矩Ferret算法对裂缝长度和宽度进行测量分析。(4)为了验证本文检测算法的性能,分别从准确率、完整率和F-测度三个方面进行评价。实验中,针对背景较均匀、对比度低及有块状阴影等复杂路面裂缝图像,将本文算法分别与大津阈值分割、Canny边缘检测、最小生成树算法、K均值聚类算法和模糊C均值聚类算法进行对比分析。实验结果表明本文算法检测的准确率、完整率和F-测度值均较高,进一步验证了本文算法具有较好的适用性。
[Abstract]:With the rapid development of highway traffic, pavement condition detection and maintenance has become the primary task of highway construction in China. Crack is one of the most important indexes to measure the pavement quality, so the detection of pavement crack by digital image processing technology has become a hot spot in this field. In the actual detection process, because of the complexity of the road surface, there are some interference factors such as oil pollution, shadow, uneven illumination and random noise in the collected road surface images. In such cases, the existing crack detection methods have the problems of misjudgment and missed detection, which can not meet the needs of detection, can not obtain more accurate and comprehensive crack information, and can not effectively maintain and manage the pavement in time. In view of the above problems, this paper mainly studies the crack detection of complex pavement from the following four aspects: road image enhancement, crack area extraction, breakpoint connection and parameter calculation. The evaluation of crack detection algorithm. (1) considering that the image of complex pavement crack has many features such as interference noise, shadow and uneven illumination, this paper proposes an enhancement algorithm of pavement crack image based on wavelet analysis. The low frequency component is enhanced by nonlinear transformation, and the high frequency component is de-noised by wavelet threshold to suppress the noise information of the high frequency part, and then the enhanced image is obtained by wavelet reconstruction. The experimental results show that the proposed algorithm not only enhances the contrast of pavement images, but also can keep the edge details of cracks as much as possible while suppressing noise. (2) aiming at the background blur, In this paper, a method of pavement crack extraction based on Legendre moment and fractional integral is proposed. Firstly, the best connected region similar to the reference image is found by Legendre moment. That is to find the best fractional integral order; Then a fractional integral mask of the optimal order is used to process the image, which reduces the gray level of the pixels in the image. Finally, the histogram of the image processed by the fractional integral operator is calculated, and the optimal threshold is determined according to the histogram to extract the crack information. This method makes full use of the property of fractional integral, considers the spatial distribution of pixels, increases the uniformity of image, and not only removes a large number of noise interference points, Moreover, the fracture area is extracted completely. (3) for the discontinuity and fracture of the extracted cracks, this paper adopts a fracture connection algorithm based on the region search to connect the fracture breakpoints. The crack neighborhood is searched according to the principle of depth first, and the breakpoint is removed according to the principle of connectivity. The length and width of the connected cracks are measured, and the crack length and width are measured and analyzed by skeleton extraction method and second moment Ferret algorithm. (4) in order to verify the performance of the proposed algorithm, the accuracy of the proposed algorithm is analyzed. Integrity rate and F-measure were evaluated in three aspects. In the experiment, aiming at the images of complex pavement cracks, such as uniform background, low contrast and block shadow, the algorithm of this paper is segmented with Otsu threshold, Canny edge detection, minimum spanning tree algorithm, etc. K-means clustering algorithm and fuzzy C-means clustering algorithm are compared and analyzed. The experimental results show that the accuracy, integrity rate and F- measure value of the proposed algorithm are high, which further verifies the applicability of the proposed algorithm.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U418.6;TP391.41

【参考文献】