手持式自定位增材制造三维形貌检测技术研究

发布时间：2018-11-20 21:29

【摘要】：增材制造技术是一种通过逐层累加来实现成形的制造技术。由于每次堆敷都存在很多不可控的变量,如散热条件、热积累等因素,导致堆敷尺寸容易出现偏差。所以,如何提高增材制造产品精度是当前增材制造面临的主要研究问题之一。本文开发了一套具有手持式、自定位特点的三维形貌扫描系统,并建立了闭环的增材制造系统,提高了增材制造产品的成形精度。首先,本文建立了手持式自定位的三维扫描实验系统,使用张正友标定法标定了相机的内外参数,并提出了一种自动高效标定结构光平面方程的方法。利用双目视觉原理计算标定点的三维坐标,精度达到了0.2mm,根据结构光三角法恢复了单帧扫描图像,得到了单帧图像被扫描物体结构光上点的三维信息,精度达到了0.1mm。优化了圆形标定点中心坐标提取的算法,实现了相同帧左右目相机图像坐标系下标定点的匹配,及相邻帧图像相机坐标系下标定点的立体匹配,使用标定点信息将不同帧图像的数据拼接起来,完成数据拼接,获取整体点云。根据上述原理,使用C++语言实现了具有在线采集处理图像,显示获取三维点云功能的软件。扫描精度在0.5mm以下,在保证点云之间最大间距为0.5mm的情况下,扫描速度可以达到15cm/min。针对电弧增材制造一般都是在平板上进行的现象,提出了一种投影到平板上均匀化点云的算法,有效的剔除了重复信息,在保证有效信息不丢失的情况下,大幅度减少了点云的数量,提高了点云处理的效率。结合点云统计滤波,高斯滤波和基于Delaunay三角形的贪婪投影三角化算法,对点云进行后处理并利用点云重建成曲面。使用C++语言实现了上述功能,利用OpenGL三维显示技术,在MFC环境下实现了模型的动态三维显示。将三维模型保存成STL格式,为其他三维模型处理软件如CAD、SolidWorks等留出了接口。最后,分别对电弧增材制造过程中单层多道及多层多道堆敷情况,提出了堆敷道特征信息提取算法。对每层堆敷形貌都进行了三维形貌恢复,特征信息提取并根据检测到的尺寸信息,改进了下一层堆敷参数和路径,建立了闭环增材制造系统,提高了增材制造产品的成形质量。
[Abstract]:Adding material manufacturing technology is a kind of manufacturing technology which can realize forming by layer-by-layer accumulation. Because there are a lot of uncontrollable variables such as heat dissipation condition and heat accumulation, it is easy to deviate from the size of stack coating. Therefore, how to improve the precision of material-increasing manufacturing is one of the main research problems. In this paper, a three-dimensional scanning system with the characteristics of hand-held and self-positioning has been developed, and a closed-loop material increasing manufacturing system has been established to improve the forming accuracy of the products. Firstly, a three dimensional scanning experiment system of hand-held self-positioning is established. The internal and external parameters of the camera are calibrated by Zhang Zhengyou calibration method, and a method of automatically and efficiently calibrating the structural light plane equation is proposed. Using the binocular vision principle to calculate the 3D coordinates of the scalar points, the accuracy reaches 0.2mm. according to the structure-light triangulation method, the single-frame scanning image is restored, and the three-dimensional information of the points on the structural light of the scanned object is obtained. The precision is 0.1 mm. This paper optimizes the algorithm of extracting the center coordinate of circular frame, realizes the matching of the subscript point of the same frame left and right camera image coordinate system, and the stereo matching of the next frame image camera coordinate system. The data of different frame images are stitched up with the information of mark and fixed point to complete the data mosaic and obtain the whole point cloud. According to the above principle, the software with the function of on-line collecting and processing images and displaying and acquiring 3D point cloud is realized by using C language. The scanning accuracy is below 0.5mm and the scanning speed can reach 15 cm / min when the maximum distance between point clouds is 0.5mm. In view of the phenomenon that arc augmentation manufacturing is generally carried out on a flat plate, an algorithm for projecting the point cloud homogeneously on the plate is proposed, which effectively removes the repeated information and ensures that the effective information is not lost. It greatly reduces the number of point clouds and improves the efficiency of point cloud processing. Combining statistical filtering with point cloud, Gao Si filter and greedy projection triangulation algorithm based on Delaunay triangle, point cloud is processed and reconstructed into curved surface. C language is used to realize the above functions and OpenGL 3D display technology is used to realize the dynamic 3D display of the model in the MFC environment. The 3D model is saved into STL format, and the interface is provided for other 3D model processing software such as CAD,SolidWorks. Finally, an algorithm for extracting characteristic information of single layer multi-channel and multi-layer multi-channel stack is proposed. The 3D morphology of each layer was restored, and the feature information was extracted. According to the detected size information, the parameters and path of the next layer were improved, the closed-loop material increasing manufacturing system was established, and the forming quality of the product was improved.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG806

【参考文献】