基于异源图像传感器的电熔增材构件表面三维重构研究

发布时间：2018-08-09 15:28

【摘要】：电弧增材制造技术(Wire Arc Additive Manufacture,WA-AM)是通过焊接电弧瞬间产生高温将金属丝材熔化,焊枪按照预设的焊接路径将熔融态金属逐层堆积,从而实现金属零件的成形目的,具备快速成形大尺寸零件的能力。WA-AM成形过程中,对热输入量有直接影响的因素有焊接电流电压、并行加工的焊枪数量及焊枪移动速度等,这些因素均会对金属熔池的形态、构件温场的分布和金属凝固的时间产生影响,是制造过程中金属构件成形稳定性、形变量和表面质量的重要因素影响。本文以多热源大型电熔增材制造生产过程中构件表面的三维重建为研究对象,以准确获取比较复杂或者有较大深度变化部件的三维表面尺寸和温度数据为目标,在对计算机视觉中双目立体视觉原理模型、标定、双目矫正、立体匹配算法和三维重建研究的基础上,设计了可见光摄像机和红外热像仪联合标定的方法以及红外热图像三维重建的方法。在Intel计算机视觉库OpenCV提供的视觉开发平台下,首先进行了可见光双目立体视觉的研究和实现,包括软硬件平台的搭建、摄像机的标定、双目矫正、双目立体匹配、三维重建以及重建结果精度分析,并结合MATLAB进行对比分析,得到的标定精度重投影误差不超过0.5个像素,三维重建误差约为4mm;然后此基础上将可见光双目立体与热像仪相结合,搭建了三目视觉系统以实现红外图像的三维重建;针对红外热像仪成像特点,设计了一种基于可见光双目立体视觉和热像仪三目摄像机联合标定的方法,该标定方法过程简单、操作较方便,重投影误差不超过1个像素,能够满足实际的应用需求;最后根据联合标定得到的投影矩阵等参数,将可见光双目图像的三维重建点云投影到红外热像仪成像平面,完成红外图像的三维重建,重建误差约为30mm。
[Abstract]:(Wire Arc Additive manufacture WA-AM) is to melt the metal wire at high temperature through the welding arc, and the welding torch piles the molten metal layer by layer according to the preset welding path, so as to realize the forming purpose of the metal parts. In the process of rapid prototyping large size parts, the factors that have direct influence on the heat input are welding current and voltage, the number of welding torch processed in parallel and the moving speed of welding torch. All these factors will affect the shape of metal melting pool. The distribution of temperature field and the time of solidification are the important factors affecting the forming stability, shape variables and surface quality of metal components. In this paper, the 3D reconstruction of the component surface is studied in the process of manufacturing the large fused material with multiple heat sources. The objective of this paper is to accurately obtain the 3D surface size and temperature data of the more complicated or the parts with large depth variation. On the basis of the research of binocular stereo vision principle model, calibration, binocular correction, stereo matching algorithm and 3D reconstruction, The calibration method of visible light camera and infrared thermal imager and the method of 3D reconstruction of infrared thermal image are designed. Based on the visual development platform provided by Intel computer vision library OpenCV, the research and implementation of binocular stereo vision are first carried out, including the construction of hardware and software platform, camera calibration, binocular correction and binocular stereo matching. The accuracy of 3D reconstruction and reconstruction results are analyzed and compared with MATLAB. The calibration accuracy reprojection error is less than 0.5 pixels, and the 3D reconstruction error is about 4 mm. Then, the binocular stereo of visible light is combined with the thermal imager. A trinocular vision system is set up to realize the 3D reconstruction of infrared image. According to the imaging characteristics of infrared thermal imager, a combined calibration method based on the binocular stereo vision of visible light and the trinocular camera of thermal imager is designed, and the calibration process is simple. The reprojection error is less than 1 pixel, which can meet the practical application requirements. Finally, according to the projection matrix and other parameters of joint calibration, The 3D reconstruction point cloud of the binocular image of visible light is projected to the imaging plane of the infrared thermal imager, and the 3D reconstruction of the infrared image is completed. The reconstruction error is about 30 mm.
【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】：TG444;TP391.41

【参考文献】