基于混合群智能优化的机器人立体视觉标定

准确的立体视觉模型是机器人高精密视觉定位的基础,而传统的单一非线性优化算法难以实现稳定和高精度的机器人立体视觉标定。结合遗传算法全局搜索能力强和粒子群算法局部搜索能力强的特点,提出了一种基于混合群智能优化的机器人立体视觉三步标定方法。针对非线性视觉模型,标定第一步和第二步分别对两个摄像机模型单独作线性初值求解和初次非线性优化,第三步对双目立体视觉模型作联合非线性优化,直接线性变换、遗传算法、粒子群算法分别作用于标定的三个步骤,每一步计算的结果被用作下一步的初始化。仿真试验分析与实际试验结果表明,相对于传统的优化标定方法和使用单一群智能优化算法的标定方法,该方法在噪声环境下具有更高的准确性和鲁棒性,能够更好满足机器人精密视觉操作的需求。

Robot Stereo Vision Calibration Based on Hybrid Swarm Intelligent Optimization

Accurate stereo vision model is the basis of robot high-precision visual positioning,however,it is difficult for the traditional or single non-linear optimization algorithm to achieve stable and high-precision calibration for robot stereo vision.Combining with strong global search ability of genetic algorithm(GA) and strong local search ability of particle swarm optimization(PSO),a three-step robot stereo vision calibration method based on hybrid swarm intelligent optimization is proposed.The calibration method is based on robot binary vision nonlinear model,linear initial values and first nonlinear optimized values of single camera models can be obtained in the first and the second steps individually,and the nonlinear optimization of stereo vision model are taken in the third step.Direct linear transformation,GA and PSO are individually used in three stages,and the result of every stage are used to initialize its next stage.Simulation analysis and actual experimental results indicate that this calibration method can work more accurately and robustly in noise environment,compared with other calibration methods using traditional optimization or single swarm intelligent optimization,and can better meet the requirements of robot sophisticated visual operation.