基于神经网络的视觉伺服机器人摄像机标定

Camera calibration for visual servoing robots based on the neural network

采用可调节点径向基函数神经网络实现视觉伺服机器人摄像机标定。首先将基于leave-one-out准则的orthogonalforward selection算法扩展到多入多出的RBF网络,建立摄像机标定的RBF网络模型。通过应用卡内基—梅隆大学标定图像实验室提供的标定数据进行仿真试验,验证此方法的有效性。由于OFS-LOO算法可构造出具有稀疏隐层节点的RBF网络,使网络具有较好的泛化推广能力,同时RBF网络为局部逼近网络,因此,此标定方法具有较高的标定精度和较强的标定实时性,适用于视觉伺服的摄像机标定。

In this paper,a new approach based on the radial basis function network for solving the camera calibration problem is proposed.In this approach,an orthogonal forward selection algorithm based on the leave-one-out criterion is applied to the multi-input and multi-output RBF network,and we build the RBF network modle of the camera calibration.The results of the simulation of the experiment indicates that the methed is valid.Because the method has the features of the higher precision and the real time,it is the same with the camera calibration of the visual servoing.