基于计算机视觉的路面破损检测与识别的研究

发布时间：2019-03-31 18:27

【摘要】：随着我国高等级公路建设的快速发展,公路的检测和维护工作在国家经济建设以及民生建设中的作用受到越来越多的重视。目前,传统的人工检测已不能满足公路快节奏发展的要求,因此,路面破损自动检测技术的研究变得尤为重要。近年来,随着计算机技术的迅猛发展,基于计算机视觉的路面破损检测系统已经应用于公路破损检测和养护领域。本文针对路面破损检测算法中存在的热点、难点进行了重点研究。首先,针对检测系统采集到的图像存在光照不均匀、噪声干扰严重等问题,本文提出对其进行预处理,图像去噪采用中值滤波处理,并将结果与均值滤波、高斯滤波效果比较,实验结果表明中值滤波在去除噪声点时可以取得较好效果。其次,对预处理后的图像进行边缘检测和区域填充。本文将双树复小波变换和直方图方向梯度计算相结合,提出一种基于双树复小波变换的路面裂缝检测算法。该算法用双树复小波变换对路面裂缝图像进行子带分解,对各子带图像进行直方图方向梯度矩阵计算,阈值化后确定裂缝边缘。实验结果表明,与传统边缘检测算法相比,该算法目标识别度高、抗干扰能力强及准确率高。确定裂缝边缘后进行区域填充处理,完全分割出裂缝区域。最后,对特征提取和分类识别进行了研究。本文对阈值化后的路面裂缝图像进行特征提取,根据提取出的特征设计了支持向量机分类器,对路面图像进行识别分类。通过多幅路面裂缝图像实验证明,该分类器可以有效地对路面裂缝图像进行分类处理。
[Abstract]:With the rapid development of high-grade highway construction in China, more and more attention has been paid to the role of highway detection and maintenance in the national economic construction and the construction of people's livelihood. At present, the traditional manual detection can no longer meet the requirements of the rapid development of highway. Therefore, the research on automatic detection technology of pavement damage becomes more and more important. In recent years, with the rapid development of computer technology, pavement damage detection system based on computer vision has been applied in the field of highway damage detection and maintenance. This paper focuses on the hot spot and difficult point of pavement damage detection algorithm. Firstly, aiming at the problems of uneven illumination and serious noise interference in the image collected by the detection system, this paper proposes to pre-process the image denoising using median filter, and compares the result with the mean filter and Gao Si filtering effect. The experimental results show that the median filter can achieve better results in removing noise points. Secondly, the edge detection and region filling of the pre-processed image are carried out. In this paper, a pavement crack detection algorithm based on double-tree complex wavelet transform is proposed by combining double-tree complex wavelet transform with histogram direction gradient calculation. The algorithm uses the double-tree complex wavelet transform to decompose the pavement crack image, calculates the histogram direction gradient matrix of each sub-band image, and determines the crack edge after thresholding. The experimental results show that compared with the traditional edge detection algorithm, the proposed algorithm has the advantages of high target recognition, strong anti-jamming ability and high accuracy. After determining the edge of the fracture, the region filling is carried out, and the crack area is completely divided out. Finally, feature extraction and classification recognition are studied. This paper carries on the feature extraction to the thresholding pavement crack image, designs the support vector machine classifier according to the extracted feature, carries on the recognition and classification to the road surface image. Experiments on several pavement crack images show that the classifier can classify pavement crack images effectively.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U418.6;TP391.41

【参考文献】