Abstract: To achieve high precision,low cost and fast calibration in the industrial sites,a line-scan camera calibration method based on Bald Eagle Search (BES) algorithm was proposed.A two-dimensional checkerboard was used as the calibration object.The feature point information was obtained by utilizing the motion of the conveyor belt.To obtain the initial values of the calibration parameters,the angle between the pixel array direction and the motion direction was first narrowed down through preliminary calibration,making it easier to simplify the imaging model reasonably.Drareni method was used to estimate the initial values of the calibration parameters.To calculate the accurate values of the calibration parameters,the motion vector was added to the calibration parameters,and the BES algorithm was applied.This method avoided the limitation of Drareni's approach,which required pixel arrays to be strictly perpendicular to the direction of motion,and did not require any additional auxiliary equipment.Experimental results showed that the proposed method could effectively calibrate even when the pixel array direction was not perpendicular to the motion direction,with a root mean square error of 0.294 pixels.This method had higher calibration accuracy and ease of application,making it suitable for meeting the calibration accuracy requirements of industrial visual measurement systems.

Key words: line-scan camera, camera calibration, parameter rotation, evolutionary algorithm, industrial site

摘要： 为了实现工业现场中的高精度、低成本、快速标定,结合某视觉测量系统实际需求,提出了一种基于秃鹰搜索算法(BES)的线阵相机标定方法。该方法以二维棋盘格作为标定物,利用输送带运动获取特征点信息。为获得标定参数初值,首先通过初步标定缩小像元阵列方向与运动方向间的夹角,以便于对成像模型进行合理简化,然后利用Drareni法估计标定参数初值。基于此,将运动向量添加到标定参数中,并应用BES算法计算得到标定参数精确值,从而避免了Drareni法对于像元阵列方向严格垂直于运动方向的限定,且无需其他辅助设备。实验结果表明,该方法在像元阵列方向与运动方向不垂直时仍能有效标定,重投影均方根误差为0.294 pixel,标定精度更高且便于应用,能够满足工业视觉测量系统的标定精度要求。

关键词: 线阵相机, 相机标定, 参数轮换法, 进化算法, 工业现场