基于粒子群遗传算法的BP神经网络摄像机标定

摘要

摄像机标定是从二维图像提取三维空间信息的关键步骤，标定的精度直接关系到三维重构结果的逼真程度。为了有效解决传统摄像机标定算法中的多参数、计算费时费力等问题，提高摄像机标定的精度和速度，将粒子群遗传算法（particle swarm optimization genetic algorithm，PSO-GA）应用于摄像机标定中。对参数进行粒子群算法优化后，再使用遗传算法中的选择、交叉和变异等操作进行参数优化，以实现粒子群算法与遗传算法的融合。结合后的算法全局搜索能力较强，收敛速度更快，优化能力与鲁棒性得以提高。同时，基于神经网络的摄像机标定方法所能覆盖的标定空间十分有限，提出了一种采用粒子群遗传算法优化BP神经网络的摄像机标定方法，以解决传统摄像机标定方法难以解决的问题。实验数据表明，基于粒子群遗传算法的BP神经网络标定是一种可行的方法，标定精度高，收敛速度快，泛化能力强。%Camera calibration is a key step for extracting three-dimensional information from two-dimensional image, which directly determines the accuracy of 3D reconstruction. In order to solve the problem of multiple parameters, reduce the computational cost, promote the accuracy and speed of camera calibration, this paper firstly applies particle swarm optimization genetic algorithm (PSO-GA) to camera calibration. The initial parameters of the genetic algo-rithm are optimized by particle swarm optimization. After that, the parameters are optimized by the selection, cross-over and mutation operations of genetic algorithm, which can realize the integration of particle swarm optimization and genetic algorithm. The resulting algorithm has stronger global search ability, faster convergence speed, better optimization ability and robustness. At the same time, the camera calibration method based on neural network just can cover very limited calibration space, this paper proposes a new camera calibration method using particle swarm optimization genetic algorithm to optimize the BP neural network, in order to solve the problem that the traditional camera calibration method is difficult to solve. The experimental data show that the BP neural network calibration based on particle swarm optimization genetic algorithm is a feasible method, which has high calibration precision, fast convergence speed and strong generalization ability.