【摘要】 为了能够实现线结构光传感器的智能标定,构建了智能标定系统。对所述系统的标定原理、光条中心线提取算法、基于GA-RBF的神经网络训练方法等算法进行研究。首先,根据异形块的特征介绍了针对异形块的标定原理。然后,利用Hessian矩阵提取出激光投射器所投出光平面与异形块相交的光平面上的棱线图;然后,利用GA-RBF算法对像素坐标与空间坐标之间的关系进行求解,即基于遗传神经网络的径向神经网络训练方法求解矩阵间的变换关系;最后,利用提出的坐标与三坐标所测量的数据进行比较,验证标定精度。实验结果表明:基于异形块的GA-RBF方法对线结构光传感器标定精度较高,满足实际应用的需要。更多还原

【Abstract】 In order to realize intelligent calibration of line structured light sensor,an intelligent calibration system is constructed.This calibration system carries out the extracting algorithm to calibration principle and the center line of light strip,and some neural network training methods based on GA-RBF are studied.Firstly,the calibration principle of the special-shaped block is introduced according to the characteristics of the special-shaped block.Then,the Hessian matrix is used to extract the ridgeline diagram of the laser projector on the light plane where the light plane intersects the special-shaped block.Then,GA-RBF algorithm is used to solve the relationship between pixel coordinates and spatial coordinates,that is,the radial neural network training method based on genetic neural network solves the transformation relationship between matrices.Finally,the calibration accuracy is verified by comparing the proposed coordinates with the measured data in three coordinates.The experimental results show that the calibration accuracy of the GA-RBF method based on the special-shaped block is higher than that of the line structured light sensor,which meets the needs of practical application.更多还原