改进型椭圆拟合的特征点质心分步提取法

Improved elliptic fitting multi-step extraction method of feature point centroid

为改善远距离视觉姿态测量系统中测量距离不同造成的质心提取误差,研究了一种基于改进型椭圆拟合的质心分步提取法。首先,对预处理后的图像采用加权灰度质心法快速获取特征点整像素,实现质心粗定位;然后,针对存在重合交叉的边缘,通过8向链码筛选获取光滑连续单像素边缘,对边缘信息归一化后进行椭圆拟合,运用3σ准则剔除拟合粗差点后,输出椭圆中心坐标;最后,通过反归一化获得质心亚像素坐标,实现质心精确定位。仿真结果表明:该方法能有效抑制图像噪声,质心提取精度优于0.05 pixel;将其应用于姿态测量系统中,测量距离为3~15 m时,以精密转台为角度基准,该方法相较于传统椭圆拟合法,最高可提升47%的测角精度。

In order to reduce the centroid extraction error caused by the different measuring distance in the long-distance visual attitude measurement system,multi-step extraction method based on improved ellipse fitting for centroid was studied.First of all,the weighted gray centroid method was used after preprocessing the image to obtain the whole pixel of feature points,which was the coarse location of centroid.Secondly,for edges with overlapping intersections,8-direction chain code scree-ning was utilized to get smooth continuous single-pixel edges,and then ellipse fitting was performed after normalization of edge information.Subsequently,the 3σcriterion was used to eliminate the gross error of fitting points,and the ellipse center coordinates were got.Finally,the sub-pixel coordinates of the centroid was obtained by de-normalization to achieve accurate positioning of the centroid.The simulation results show that the proposed method can effectively suppress image noise,and the accuracy of centroid extraction is better than 0.05 pixel.Meanwhile,the proposed method is applied in the attitude measurement system,taking the precision turntable as the angle reference.The experiment results indicate that the angular measurement accuracy can be raised up to 47%compared with the traditional elliptic fitting method within the range of 3~15 m.