Abstract: Aiming at the problem that the traditional edge-detection-based technology of hole recognition and location was not applied to grid-like surface,a template-matching-based algorithm for detecting the hole position and normal was proposed.The principle of measuring 3D point cloud by laser-line profile sensor was introduced.An improved median filter operator was used to remove outliers at the edge of the hole,and the 3D space point cloud was converted into a 2D plane point cloud by binarization.The hole was identified and located in 2D plane through template matching and division of Regions of Interest (ROI),and the location and normal of the hole were calculated indirectly based on the point cloud of hole neighborhood.In the template matching algorithm,the template of the hole was adaptively constructed according to the point cloud characteristics and the given aperture,which ensured the applicability of the algorithm on inclined planes and small curvature surfaces.The effectiveness and accuracy of the proposed algorithm were verified by experiments,which showed that the hole measurement accuracy was better than 0.12mm.

Key words: 2D laser scanning, template matching, template construction, hole position and normal detection, regions of interest

摘要： 传统的基于边缘检测的孔位识别定位技术无法适用于网格状表面,因此提出一种基于模板匹配的孔位与法矢检测算法。给出利用2D激光采集被测表面三维点云的原理,采用改进型中值滤波算子去除孔边缘离群点,通过二值化将三维空间点云转换成二维平面点云,在二维平面内运用模板匹配方法与感兴趣区域(ROI)进行孔的识别与定位,并依据孔邻域点云间接计算出孔的位置和法矢。在模板匹配算法中,根据点云特征与给定的孔径自适应构造出孔的模板,保证了算法在倾斜平面和小曲率曲面上的适用性。通过实验对所提算法的有效性和精度进行了验证,实验结果表明孔位测量精度优于0-12 mm。

关键词: 2D激光扫描, 模板匹配, 模板构造, 孔位与法矢检测, 感兴趣区域