测绘学报

• 学术论文 • 上一篇    下一篇

目标轮廓直线特征的高精度提取

李彩林1,郭宝云2   

  1. 1. 武汉大学遥感信息工程学院
    2. 武汉大学
  • 收稿日期:2009-12-03 修回日期:2010-07-20 出版日期:2011-02-25 发布日期:2011-02-25
  • 通讯作者: 李彩林

The High-accurate extraction of line features of object contour

  • Received:2009-12-03 Revised:2010-07-20 Online:2011-02-25 Published:2011-02-25

摘要: 针对小型物体高精度量测的要求,本文提出了一种结合精确单点定位方法和基于分裂的最小距离误差线段逼近法的高精度轮廓直线特征提取算法。通过利用直线特征精确逼近目标的轮廓,搜索相邻直线段并计算交点,从而实现轮廓分割点的精确定位,进而为基于轮廓的三维量测与重建提供高精度的直线特征数据。与其它方法(如最小二乘模板匹配、抛物线拟合等)相比,本文提出的算法对轮廓初始值的要求较低,同时能够获得更好的精度。利用实际影像数据进行实验,对该算法提取的精度进行了深入分析,并将其成功的应用于钻石的三维视觉测量系统中,取得了良好的效果。实验结果表明,该方法具有快速、稳定、精度高等特点,还可推广到其他应用领域。

Abstract: A high-accurate extraction algorithm of line features of object contour is proposed in this paper, which combines the method of exact positioning of single point and the algorithm of line segment approximation with minimum-distance error. First object contour can be approximated by linear feature. Then we search the adjacent line segment and compute the intersection point between them. At last, we realized the exact positioning of contour cutpoint and provided exact line segment data for 3D measuring and reconstruction based on contour. In contrast to other approaches, the algorithm proposed in this paper has not need good initial value of contour and can get better extraction result. The accuracy of the algorithm is analyzed on real image data. The algorithm has been applied in 3D inspection of diamond, line feature extraction of nuts and industrial sheetmetal parts. The experiment result demonstrates that the algorithm is feasible, correct, stable and exact and can also be extended to other application areas.