摄像机镜头畸变的鲁棒校正方法

Robust correction method for camera lens distortion

针对计算机视觉中的镜头畸变问题,设计一种鲁棒的校正方法.该方法基于空间直线的成像特性来定义畸变测度,通过非线性优化完成畸变校正.采用微粒群全局优化算法,将传统优化方法、标准微粒群算法和基于不同策略的微粒群算法的性能进行对比.实验结果表明,带变异算子基于对位学习的微粒群算法具有较强的鲁棒性,在低噪声下,微粒群算法的校正性能优于传统算法.最后通过不同畸变程度的校正实例验证了所提出方法的有效性.

For the problem of lens distortion in computer vision, a robust correction method is proposed. The proposed method optimizes the distortion measurement defined by the imaging properties of space lines to correct distortion. Particle swarm optimization algorithms are used, and the comparison is analyzed between the traditional optimization algorithm, standard particle swarm algorithm and other improved particle swarm algorithms based on different strategies. Experimental results show that the opposition-based particle swarm algorithm with a mutation operator has strong robustness, the performance of particle swarm algorithms is better than the traditional algorithm in the low-level noise situation, and different degrees of lens distortion verify the effectiveness of this method.