【作者】 杨雪荣；

【导师】 张湘伟；

【作者基本信息】 广东工业大学 ， 机械制造及其自动化， 2013， 博士

【摘要】 逆向工程(Reverse Engineering, RE)技术已成为支持复杂产品设计,缩短产品开发周期、提高企业创新能力的有效工具,是先进设计和制造的重要技术手段。实物表面数字化、数据处理、模型重建是逆向工程的关键技术。随着逆向工程技术的发展,单一传感器已不能满足实物表面数字化对精度和效率越来越高的要求,同时对复杂零件,单一传感器通常难以获得模型的完整信息。多传感器集成测量对实现高效率、高精度、完整测量的逆向工程提供了新思路,但同时也对后续的数据处理、数据融合、模型重建提出了新的要求。论文在系统地分析了国内、外关于多传感器集成测量理论与应用研究的基础上,结合线结构光视觉测量与三坐标测量机(Three Coordinate Measuring Machine, CMM)接触式测量各自的特点和优势,将其有机集成,提出了一种CMM与线结构光视觉传感器集成的逆向工程系统。论文就集成测量、数据处理及数据融合等关键技术展开了研究,主要内容包括：1、提出一种CMM与线结构光视觉传感器集成测量方案,对复杂产品采用分而治之的测量方法,由CMM驱动线结构光传感器以自适应扫描方式获取自由曲面初始数据,在视觉盲区及测量误差大的区域用CMM测量少数点对初始曲面进行修正,提高曲面建模精度。用CMM测量规则特征,提取其数学参数。该方法充分利用了光学测量效率高和CMM测量精度高的优点,扬长避短,实现高效、高精度的完整测量。2、建立集成系统中线结构光三维视觉测量数学模型,基于Zhang摄像机标定方法,考虑切向畸变,采用非线性全局优化方法实现任意位姿平面靶标的摄像机高精度标定。采用基于交比不变性原理和任意位姿平面靶标相结合的方法,获得光平面上任意多个特征点,实现对视觉传感器结构参数的标定。3、提出基于可变形模型的光条中心提取方法。在光条骨架处设置B样条曲线作为变形初始模型,将梯度矢量流力场(Gradient Vector Flow, GVF)作为外力,对B样条可变形模型的能量泛函极值问题进行求解,实现光条纹中心的提取。该方法不仅利用了光条的骨架几何信息而且计算整个光条能量,可有效地抑制噪声、分枝的影响,具有较强的光条纹断线修补能力。提取的光条中心具有单一连通性、亚像素级精度,而且具有和测量曲面一致的光顺性。同时,逆向建模时将二维B样条曲线转化为空间B样条曲线,可避免存储大量“点云”数据及后续散乱点云复杂处理,简化了曲面重建工作,可实现逆向工程数字化、数据处理及模型重建过程的有机结合,提高逆向工程的效率。4、提出线结构光扫描步长自适应规划的两步法,根据辅扫描线的曲率变化确定主扫描步长。针对辅扫描曲线采样问题,提出基于角度的曲线采样方法,可实现采样点的分布与曲线的弯曲程度保持一致,且算法简单。提出基于角度和弧长相结合的曲线采样方法,在曲率较小区域结合弧长法补充采样点,满足逆向工程采样完整性要求。对未知自由曲面实现了线结构光扫描测量时步长随曲率变化而自适应变化,可明显减少扫描线数量,提高测量效率和测量精度。5、提出基于曲面变形和混合建模技术的多传感器测量数据融合方法。对自由曲而部分,将线结构光测量的曲面作为初始曲面,CMM测量的少量高精度点转化为点约束形式,通过基于ACIS平台开发的变形造型软件,用能量优化法求解具有最小能量的变形曲面,实现用CMM少量的高精度数据点修正初始曲面,提高模型重构的精度,且具有良好的保形性。该方法避免了数据融合过程中由于不同种传感器测量数据密度、精度相差悬殊造成的数据难以对齐的难题。对规则特征部分,采用正逆向混合建模方法实现几何特征和自由曲面的数据融合,完成CAD模型重建。更多还原

【Abstract】 Reverse Engineering (RE) has become an effective tool for supporting complex product design, shortening product development cycles and improving innovation ability of the enterprise. It’s an important technical method to the advanced design and manufacture. Digitization, data processing and model reconstruction are the key technologies of the RE. With the development of RE, the single sensor can not meet the digitization increasing demands of the accuracy and efficiency any more. In the meantime, it’s also difficult for single sensor to obtain complete model information of the complex parts. Multiple-sensor integrated measurement provides a new idea for achieving high efficiency, high precision and complete measurement of RE. However, integration measurement present new requirements to the subsequent data processing, data merging and model reconstruction.Combined with respectively advantages and features between the line structured light vision measurement and three coordinate measuring machine (CMM), a new multiple-sensor integrated RE system is proposed, on the basis of the systematic analysis for the theory and application of the domestic and foreign multiple-sensor integrated measurement. The key technologies such as integrated measurement, data processing and data merging are researched in this thesis. The main contents are including:1. A new CMM and line structured light vision sensor integrated measuring program is proposed. Adopting the divide-and-rule method is applied to measure complex products, using CMM to drive line structured light sensor in order to obtain surface initial data. In the areas where vision measurement has great error or the blind areas, CMM measure fewer points in order to correct the initial surface. For the rule features, CMM measurement is used to get design parameters. This method makes full use of the rapidity of optical measuring system and the high accuracy of CMM, to realize the high-efficiency, high-accuracy and complete measurement.2. The mathematical model of the line structured light vision measuring is established. On the basis of Zhang’s camera calibration method, considering the tangential distortion and using nonlinear global optimization method, all the camera parameters can be calibrated with free-position planar pattern. Using the combination of cross ratio invariance theory and free-position planar pattern to obtain the numerous feature points in a light plane, the structural parameters of the vision sensor can be calibrated.3. The light strip centre extraction method based on deformable model is proposed. B-spline curve is set on the the strip skeleton as the initial model of the deformation model. Regarding the Gradient Vector Flow (GVF) as external force, the model is deformable operated by internal force and external force. The centre of light stripe can be found when energy of deformable model is least. The proposed approach not only takes advantage of the geometry information of the light strip skeleton but also calculates the energy of the entire light, so that it can restrain the effects of noise and branching efficiently, and also has a strong ability of repairing broken light stripe. The extracted centre of the light strip has single connectivity, smoothness as the free-form surface and subpixel accuracy. Meanwhile,2D B-spline curve is converted into3D B-spline curve space based on the3D measurement model. It can avoid storing and processing large amounts of scattered "point cloud" data, and also can simplify surface reconstruction. This approach can achieve the combination of digitization, data processing and model reconstruction, which improves the efficiency of RE.4. The two-step method of line structured light scanning adaptively is proposed, which will determine the steps of main scanning based on the curvate of the auxiliary scan line and transform the freefrom surface sampling into the curve sampling.A fast and efficient sampling approach based on the angle analysis of the curve is proposed, which will realize the points adaptive distribution with the curvature of curve. When combined with arc-length method, it can avoid the problem of less sampling points in the small curvature region. The approach can make scanning step change adaptively with the curvature of curve when the line structured light scans the unknown freeform surface. It can significantly reduce the number of scan lines and improve the efficiency and accuracy of vision measurement.5. The method of multiple-sensor data merging, based on surface deformation and hybrid modeling technology, is proposed. The line structured light sensor measure the free surface of the part as the initial surface, and transforming the small number of high-precision points got by CMM to the point constraint forms, in order to correct initial surface.Through the developed deformation modeling software on the ACIS platform, the initial surface is deformed to high-precision points based on the energy optimization method. The proposed approach of data merging can not only improve the utilization of measurement information by different sensors and the accuracy of the surface reconstruction, but also make reconstruction model has good conformality. It can also avoid the problems that the date got by different kinds of sensor can not be alignment due to the disparity of data density and data accuracy. For the rules elements, the hybrid modeling method based on the feature is used to fuse rules elements measured by CMM and freeform data into the complete CAD model.更多还原