基于小波神经网络在摄像机标定中的研究

Research on Camera's Calibration Based on Wavelet Neural Network

摄像机标定是计算机视觉中一项关键环节,神经网络具有强大的非线性映射逼近能力,通过神经网络可以建立物理世界坐标和相机坐标的隐式标定。由于小波神经网络比传统BP神经网络训练从根本上避免局部最优且加快收敛速度,具有很强的学习和泛化能力,也避免网络结构盲目设计。利用小波神经网络自适应动量快速学习算法来标定相机,通过实验仿真取得良好效果。

Camera＇ s Calibration is an important part of computer vision. The neural network can implicitly calibrate between physical coordinates and Camera coordinates with strong approximation ability of nonlinear mapping. Compared with the calibration of traditional BP neural network, the wavelet neural network can effectively avoid local optimization ,blindness of network design and quick convergence rate. It has strong learning and generalization ability. It makes use of wavelet neural network of self-adaption momentum fast training algorithm to get the camera＇s internal and external parameters with good results of experimental simulation.