视觉定位技术是机器视觉当中一个重要的领域，它在移动机器人中是必 不可少的。关于视觉定位最常见的方法是先根据图像提取特征点，再根据这 些特征点去定位。然而,当场景中缺失纹理时，视觉方法通常都会失败。例如， 对于室内大多数的平面或者是工厂的很多零部件，很难提取丰富的特征点。 但是，这些场合中通常存在有显著的边缘特征，很多零部件的轮廓是规则的 圆形。因此，对这些边缘图像特征进行有效提取具有重要的应用价值。 圆形的图像是二次曲线，目前已有的方法通常是对单个的二次曲线进行 提取，还没有多个二次曲线的自动有效提取方法。本文提出了一种从图像当 中提取多个二次曲线并对每个二次曲线拟合的自动算法。该算法首先对采集 到的图像进行边缘检测，然后使用图像分割技术检测出形状为二次曲线的连 通区域，最后根据一种基于几何距离的二次曲线拟合算法拟合检测出来的连 通区域。通过实验效果图和实验数据表明，本文提出的算法成功的实现了数 字图像当中多个二次曲线的拟合；从实验效果图可以看出这个算法具有较高 的精度。

Visual localization technology is an important problem in machine vision and indispensable in mobile robot. The most common way of visual localization is to extract feature points from images and then to locate the camera based on these feature points. However, since there are lots of poorly textured scenes indoor or in the factory, it is difficult to extract the rich feature points. But, there are some salient edges. It follows that how to effectively extract these edges is important. In this paper, an algorithm for multiple conic fitting based on a geometric distance is proposed. This algorithm can extract multiple conic from images and fit every conic based on a geometric distance. The algorithm first detects the edge of the captured image, and then uses the image segmentation technique to detect the connected region whose shape is a conic. Finally, all the connected regions that are detected are fitted by a conic fitting algorithm based on a geometric distance. The experimental results show that the proposed algorithm has successfully realized the fitting of multiple conic in the digital image with a high accuracy.