改进了一种基于工业计算机断层成像(industrial computed tomography,ICT)图像与计算机辅助设计(computer aided design,CAD)模型的比对检测算法,分析工件制造误差。首先对工业CT图像用三维Otsu法进行阈值分割,并分别提取边缘面与角点特征;然后对工业CT图像角点特征与工件的CAD模型用文中研究的旋转投影法求取方向包围盒,进而实现粗配准;再结合角点特征点集和奇异值分解—迭代最近点算法进行精配准,最近点对的求取用k-d树进行加速;最后在边缘面上显示误差。实验结果表明,该方法在工件比对检测过程中,粗配准精度更高,适应性更好。整个比对检测过程更加高效,速度上有了较大的提高。
This paper improved an algorithm of analyzing the manufacture error based on comparison inspection between ICT images and the CAD model.Firstly,it segmented the ICT images by 3D Otsu threshold method,and then obtained the edge surface and corner features.Secondly,it calculated the oriented bounding boxes(OBB) of the ICT images' corner features and the work-piece's CAD model using the presented rotating projection method presented,then realized the rough registration by the two OBBs.Then the singular value decomposition and iterative closest point(SVD-ICP) algorithm were used to complete the precise registration between the CAD model and corner features of ICT images.The k-d tree was used to improve the calculation speed of searching for the closest point.Finally,it displayed the error using edge surface.The experimental results indicate that the result of rough registration in this paper is more accurate and applicable.The whole algorithm is more rapid and efficient.