摘要
提出新的有效的位姿测量算法。该方法使用共面特征点,利用投影变换中的平行性约束等仿射不变量快速求得特征点摄像机坐标系空间深度值,并作为初值求解以特征点几何约束条件建立的无约束非线性最优化目标函数,保证最终解的精确性和收敛性。搭建实验系统验证了算法的有效性,并与最小二乘法进行比较,该算法有效减少了计算过程中的迭代次数,提高了运算速度。该结果为应用单目视觉进行工业实时在线测量提供了一种新的方法。
This paper focused on the need for the improvement of the accuracy and speed of the algorithm in the process of solving the monocular vision object position and orientation.The algorithm built on the invariants of parallel relations and distance factor of the co-planar feature points of the target,and quickly calculated the depth of the feature points in the camera coordinate system.Then it established the nonlinear mathematic model according to co-planarity of the feature points,improved the accuracy of the solution.The algorithm simplified the computational model,reduced the number of iterations,and improved the speed in the solving process.The experimental results show that it has a good precision and enhances the solution speed.It provided a new method for real-time on-line measurement.
出处
《计算机应用研究》
CSCD
北大核心
2013年第6期1917-1920,共4页
Application Research of Computers
关键词
单目视觉
位姿测量
共面特征点
平行性约束
透视投影
monocular vision
pose estimation
coplanar feature points
parallel constraint
perspective projection