大型自由曲面快速柔性视觉定位测量方法研究

According to the digital assembly process characteristics of large-scale equipment, 3D profile and spatial position of work-pieces must be computed online. For free-form surfaces, the complex nature of the geometry makes the measurement solutions more difficult. A flexible visual positioning method is proposed in this project. Measuring sensors move around the large-scale equipment during the assembly process, and a multi-layer visual positioning and tracking network is formed. As different from the traditional visual network, this is called the flexible visual positioning network. This project focuses on basic theories, key technologies and construction methods of the above flexible visual positioning system aided digital assembly of large-scale equipment. The main researching contents are listed as followings: flexible visual positioning mathematical modeling, measuring coordinate transformation theory, scale change pattern of multi-sensors motion space and errors calculation, nonlinear filtering theory for motion estimation, data fusion of multiple visual sensors and IMUs, virtual simulation technique, parallel processing method of massive profile data, measuring system construction, etc. Based on the theoretical results, computer simulation is applied to correct the system model. And then, the simulating system is built in the laboratory to test the algorithm functions and measuring performances. In the end, the real flexible visual positioning system is constructed in the assembly floor to evaluate the researching work. This project aims to research the measuring method using flexible visual network aided for digital assembly of large-scale equipment, develop the measuring device with independent intellectual property, and contribute to improve the assembly quality of large-scale manufacturing devices.

大型设备精密测量与定位技术是制约其制造、装配质量的关键因素，在测量领域占有特殊重要地位。大型设备及其部件的外形多属自由曲面，采用常规测量仪器无法获取完整三维形貌信息，且量程与精度矛盾突出。本项目提出大型自由曲面快速柔性视觉定位测量方法，精确获取复杂曲面三维形貌信息或设备位姿信息。主要内容包括：⑴研究大型自由曲面测量坐标系转换理论和多传感器运动空间尺度变化规律；⑵建立柔性视觉定位测量数学模型；⑶研究运动状态估计非线性滤波方法和多传感器测量数据融合理论；⑷建立柔性视觉定位测量分布式高效计算模型；⑸构建大型自由曲面柔性视觉定位测量装置。项目拟在无控制点的测量坐标系转换、多传感器融合运动误差补偿、曲面测量海量数据并行处理等关键技术环节取得突破，以此为基础，通过计算机仿真、实验室模拟以及现场实验相结合的方式对系统性能进行验证与评价。本项目研究将为大型设备形貌或位姿在线快速测量提供新的技术途径和手段。

大型设备精密测量与定位技术是制约制造、装配质量的关键因素，在测量领域占有特殊重要地位。大型设备及其部件的外形多属自由曲面，采用常规测量仪器无法获取完整三维形貌信息，且量程与精度矛盾突出。本项目采用基于二维-三维特征相融合的快速柔性视觉定位测量方法，通过视觉测量传感器的自由移动扩大测量范围，精确获取大型复杂曲面三维形貌信息或设备位姿信息。.项目在对大型自由曲面测量坐标系转换理论、二维图像特征匹配方法、运动状态估计非线性滤波方法、基于惯性传感器件的视觉传感器姿态估计方法等进行研究的基础上，搭建了基于二维-三维信息相融合的无约束视觉测量系统，利用经过优化的双目结构光视觉传感器获取单站位三维测量数据，通过二维特征匹配和视觉传感器姿态估计数据计算测量站位间的坐标转换关系从而实现测量数据的粗拼接，之后对三维测量点云数据进行三维特征提取及迭代处理，减小拼接误差，并进一步获取大型自由曲面完整形貌信息或设备位姿信息。.项目所提出的基于二维-三维特征相融合的快速柔性视觉定位测量方法在所搭建的无约束视觉测量系统中进行了验证。在项目研究成果基础上，对视觉传感器姿态估计滤波方法和三维特征匹配方法进行深入研究，并在此基础上优化无约束视觉测量系统、进一步提高测量数据的并行处理速度，可为大型设备形貌或位姿在线快速测量提供新的技术途径和手段。