摘要
传统的计算视觉测量方法,普遍存在图像特征提取困难,非线性,求解过程复杂的问题。针对这种情况,讨论了一种基于正交优化的线性视觉测量方法,即利用霍夫变换提取图像中的直线,再利用对应直线列解线性方程,最后对解矩阵进行正交优化。它具有求解简单,容易操作的特点,且对误差具有鲁棒性。最后针对不同旋转角度进行了大量实验测量,证明了方法的可靠性。
There are a lot of difficulties in traditional vision-based measurement methods. Equations which are always non-linear and complicated have to be solved, and features are often not easy to extract. Focused on these problems, a feature measurement algorithm based on Hough translation is discussed. Corresponding lines in different pictures are extracted first, and then linear equations are given by line features, finally the result matrices are orthonormalized and optimized. This algorithm has many advantages, the equations are much simpler, and the robustness to errors is good. Its effectiveness is revealed by later experiments and data comparisons.
出处
《微计算机应用》
2007年第3期229-232,共4页
Microcomputer Applications
关键词
计算视觉
对应直线
线性方法
正交优化
lines correspondence, rotation matrix, error analysis, solution optimization
