Abstract: In order to recognize log end effectively, an ellipse fitting algorithm based on non-linear least squares principle is studied. The accuracy of ellipse fitting is largely affected by the initial value. This method calculates the distance of the target image’s boundary points to obtain the appropriate initial value. The smallest Euclidean distance between the boundary points and fitting ellipse is caculated by using least squares principle, thus the best fitting ellipse parameters of major axis and minor axis are gotten. The results show that the proposed algorithm has good fitting accuracy and applicability for recognizing log end.

Key words: non-linear least squares, ellipse fitting, log scaling, Euclidean distance

摘要： 为了更好地进行原木端面识别，研究了一种基于非线性最小二乘原理的椭圆拟合算法。椭圆拟合的精度在很大程度上受初始值的影响，该方法通过对目标图像的边界点进行距离计算，得到了适当的初始值；之后运用最小二乘原理，计算边界点到拟合椭圆之间欧式距离的最小值，确定最优拟合椭圆的长短轴参数。实验结果表明，提出的算法在原木端面的识别中，具有良好的拟合精度和适用性。

关键词: 非线性最小二乘, 椭圆拟合, 原木检尺, 欧式距离