基于双更新策略加权差分进化粒子群的双目相机标定

针对空间目标位姿测量下的相机多参数标定问题，提出基于双更新策略加权差分进化粒子群优化的相机参数标定方法。通过引入自适应判断因子来控制每一次迭代过程中加权差分进化(WDE)算法和粒子群优化(PSO)算法的调用比例，根据概率规律考虑对个体使用PSO算法或WDE算法来进行更新，并通过信息交流机制利用WDE操作得到的个体去引导PSO操作中的个体进化过程，所提出的WDEPSO算法能够保证种群个体进化的多样性和有效性，并且与相机非线性标定模型参数进行耦合，同步实现相机内外参数的组合非线性、全局连续优化，克服目标空间背景饱和光强造成的有限特征点失效引发的局部收敛问题。实验表明，文中方法优化得到的目标函数值更小，获得了较高的标定精度；利用标定参数得到的标准杆测量精度优于0.40 mm，目标大幅度角运动状态下的重构姿态误差小于0.30°，可重复性测量结果稳定。

Binocular camera calibration based on dual update strategy weighted differential evolution particle swarm optimization

Aiming at the problem of camera calibration with multi-parameters in space target pose measurement, a novel calibration method based on dual update strategy weighted differential evolution particle swarm optimization was proposed . An adaptive judgment factor was constructed to control the usage proportion of weighted differential evolution (WDE) algorithm and particle swarm optimization (PSO) algorithm in each iteration process, which allowed the PSO or WDE algorithm to update individuals according to the probability laws. Moreover, through the information exchange mechanism, the individual obtained by the WDE operation was adopted to guide the individual evolution process operated by PSO algorithm. The proposed WDEPSO algorithm could ensure the diversity and effectiveness of the individual evolution of the population, and it was coupled with the camera nonlinear calibration model and parameters. Consequently, the proposed method could simultaneously realize the combined nonlinear and global continuous optimization, which overcomed the local convergence problem caused by the limited feature points arise from the saturated light intensity of the target space ambient. Experiments suggest that the objective function value optimized by the WDEPSO method is smaller, and the proposed method obtains the higher calibration accuracy. The standard gague measurement error obtained by the calibration parameters is less than 0.40 mm, the reconstruction attitude accuracy of the target under large angular motion is better than 0.30°, as well as the repeatability measurement results are stable.