基于差分演化算法的摄像机高精度标定

High precision camera calibration based on differential evolution algorithm

摄像机标定是机器视觉中的关键问题之一。为了提高摄像机的标定精度,提出了一种基于差分演化的摄像机标定算法。采用针孔模型和二阶畸变模型作为标定模型;用单位四元数表示世界坐标系与摄像机坐标系之间的旋转矩阵以减少需要标定的参数;采用透视不变性原理计算圆形靶标的中心,得到标定点的精确三维世界坐标;将针孔模型计算出的理想图像点与畸变模型计算出的重建图像点的位置差定义为目标函数,采用改进的差分演化算法,通过种群的遗传操作和变异操作求出摄像机的内外参数。测试结果表明,基于差分演化的摄像机标定的最大误差小于0.5 pixel,其标定精度高,标定速度快,且对初始值的依赖程度小,方法具有很强的实用性。

Camera calibration is one of the key problems in machine vision. In order to improve the accuracy of it, a new algorithm for camera calibration based on differential evolution is proposed. The pinhole model and second order distortion model are used to form a calibration model. In order to reduce the number of calibration parameters, world coordinate system and the camera coordinate system rotation matrix are replaced with a unit quaternion. At the same time, circular target center is calculated by using perspective invariance principle so that the 3 D coordinates of circular target cen-ter can be measured accurately. The objective function for camera calibration is defined by the dis-tance of the ideal image point calculated by pinhole model and the reconstructed image point compu-ted by the distortion model. Then the improved differential evolution algorithm is used to calculate the internal and external parameters of the camera using genetic operation and mutation operation of the group. Test results show that the maximum error of camera calibration based on differential evo-lution is less than 0. 5 pixel. The calibration has advantageous properties such as higher precision, faster speed and less dependence on initial value, and has potential in application.