【作者】 王明辉；

【导师】 陈静；

【作者基本信息】 武汉理工大学 ， 控制理论与控制工程， 2006， 硕士

【摘要】 随着我国经济的蓬勃发展,交通运输事业也随之取得了令人瞩目的成就,其中,高速公路的总里程已经跃居世界第二。对数量如此庞大的公路资源进行有效管理和维护,必须迅速提高路面检测技术的整体水平。大型成套检测设备是检测技术水平的具体体现,它们的应用对我国交通事业的长远发展有深远影响,对整个国民经济的发展也意义重大。 本文主要对沥青路面行车性能激光检测系统中的图像处理技术展开研究,采用理论与试验手段相结合的研究方法,重点论述了路面图像分析、路面轮廓数据管理以及路面模型重构等方面的核心技术。 文中首先介绍检测系统的总体实现方案,并对主要硬件的组成情况以及软件架构作了概述。随后,着重讨论了路面图像的处理和分析技术,通过对比与综合,提出一种高精度的复合细化算法。该算法的主要步骤分两步:1)分别采用阈值分割和灰度邻域属性法消除白噪声及孤点噪声;2)使用B样条拟合重心法完成图像的细化工作。 本文还阐述了基于NetCDF格式的轮廓数据管理方案,它区别于关系数据库系统,对路面分析数据的存储有更优的效率。在方案实现上,借助多线程及缓冲区预存技术充分减小数据的存储延时以提高系统的整体性能。最后,详细论述了基于细分的Hoppe方法实现路面模型重构的原理及特点,对于如何利用OpenGL结合MFC开发模型实现重构算法,也作了简要的讨论并给出相关的主要程序。 通过模拟试验及典型的数据测试,证明了本文采用的路面图像细化算法、分析数据管理方案和路面模型重构方法的可行性和有效性,可以应用于本文所提及的检测设备及其他同类型的设备中。更多还原

【Abstract】 With the rapid development of national economy, domestic transportation industry has attained prominent achievements among which the total mileage of highway has already ranked second in the world. It is necessary to upgrade the pavement surface detecting technique generally so that such numerous highway resources are maintained properly. Large-scale complete set detecting equipments are embodiment of detecting techniques;therefore, their application has profound influence on the long-term development of communication in our country and is supposed to be significant to domestic economy.This dissertation focuses on the image processing and analyzing techniques indeveloping Laser Inspecting System for Driving Property of Asphalt Pavement(LDSDPAP). Having implemented a method combining theory with experiment, itmainly describes core techniques on pavement image analysis, analysis datamanagement and reconstruction of 3D pavement model.First, this dissertation concisely introduces the scheme of detecting system, main hardware structure and software architecture. Second, it particularizes the pavement image processing and analyzing techniques and a high precision hybrid thinning algorithm has been proposed through comparison and synthesis. The algorithm includes two major phases: 1) removing white noise and acnode noise by threshold segmentation and grayscale neighborhood property respectively;2) using B-Spline based barycenter algorithm to finish the pavement image thinning.Third, a NetCDF based analysis data management plan has been discussed. It differs from the relational database system in efficiency of storing pavement analysis data and characterizes an implementation with a minimum delay using multithreading and buffer prefetching techniques to improve the performance of whole system.Finally, there is a full discussion of concept and features of Hoppe method based on subdivision for pavement model reconstruction and an explanation of key parts about how to program reconstruction module combining OpenGL with MFC. Meanwhile, the related coding sections are also presented for demonstration.Being testified by simulation and typical data, the pavement image thinning algorithm, analysis data management plan and pavement model reconstruction method are proved to be feasible and effective. Hence, they can be applied to LDSDPAP and other similar systems.更多还原