基于一维图像序列的大尺度三维形貌动态测量新方法

In view of the completeness and comprehensiveness of 3-D shape information, 3-D shape measurement has shown its important application value and increasing prospect in fields of large-equipment operation control, facility fault detection and condition monitoring. To meet the special requirements of large scale, high dynamic and high resolution on 3-D shape measurement in these areas, by utilizing the feature of high speed relative motion between sensor and measured object, and borrowing the idea of high resolution terrain mapping in aerospace, high-speed linear CCD image sensor is introduced into large-scale spatial 3-D shape measurement. Movement informations represented by speed are imported into existing 3-D shape measurement theory. Temporal relationship between 1-D image sequence and relative motion's position/orientation is established via global clock synchronization, then real-time estimation of the relative position/orientation is achieved during measurement process. On this basis, taking motion information as a reference, the image sequence registration and fusion are further studied as well as real-time 3-D reconstruction algorithm, and the error propagation pattern in 3-D shape measurement is also analyzed systematically. A running status monitoring and fault identification system for high-speed trains is set up to verify the new method. In this research, new ideas and new technologies can be produced for future applications such as dynamic measurement and 3-D state detection of high-speed rail trains.

鉴于三维形貌数据的完整性和全面性，三维形貌信息测量在大型装备运行控制、故障检测及状态监控等领域的重要应用价值和发展前景日益凸显。针对上述领域三维形貌测量空间尺度大、高动态、高分辨率的特殊测量要求，利用传感器-被测物间高速相对运动的特点，借鉴航天领域高分辨率地形地貌测绘的思想，将高速线阵CCD图像传感器引入大尺度空间三维形貌测量。以现有三维形貌测量理论为基础，导入以速度为代表的运动信息，通过全局时钟同步建立一维图像序列与相对运动姿态间的时空关系，实现测量过程中对相对位姿的实时估计。在此基础上，深入研究基于运动信息参考的多传感器图像序列配准融合及三维形貌实时重构算法，系统分析融合运动信息后三维形貌测量误差传递规律，并搭建高速列车运行状态监控及故障识别系统，对新方法进行现场验证。本研究可为高铁列车动态测量及三维状态检测等未来应用提供新思路、新技术。

本项目针对高铁列车等大空间运动物体三维形貌测量大量程、高精度的动态测量需求，借鉴航天航空高分辨率形貌测绘思想，利用传感器-被测物间不断相对运动的特点，在现有三维视觉形貌测量理论基础上引入相对运动位姿测量信息，从理论、算法、应用三个层面系统研究基于一维图像序列的三维形貌动态测量方法，为列车动态测量及三维状态检测等未来应用提供新思路、新技术。.项目系统研究了基于一维图像序列的三维形貌动态测量原理和方法。建立了基于线阵图像严格交会的通用三维测量模型。提出了两种线扫描相机标定方法，实现了线扫描相机现场高精度、高效率标定。研究了基于硬件加速运算的图像配准和三维形貌重构方法，解决了密集形貌测量数据实时处理难题。系统分析了线扫描形貌传感器中结构、图像等因素所导致的测量误差，并进一步研究了融合运动信息后的三维形貌测量误差传递规律，在优化传感器测量精度的基础上实现了基于运动数据融合的动态误差实时补偿。构建了运动物体三维形貌连续测量验证平台及移动机器人自动扫描测量验证平台，实现了高速、高分辨率的三维形貌自动测量。.依托本项目，获得2016年度教育部技术发明一等奖（2）及天津大学沈志康奖教金资助，入选天津大学“北洋学者-青年骨干教师计划”。协助培养博士研究生3人（已毕业2人），硕士研究生8人（已毕业4人）；申请发明专利2项，授权专利3项；发表论文12篇，SCI检索8篇，EI检索3篇。