应用BP神经网络进行成像误差修正  

Correction of Camera Distortion Based on BP Neural Networks

作　　者：袁野[1] 田中旭[2] 欧宗瑛[3] 

机构地区：[1]上海交通大学图像通信与信息处理研究所,上海200030 [2]上海大学机器人中心,上海200072 [3]大连理工大学机械系,辽宁大连116024

出　　处：《计算机应用》2003年第6期109-111,共3页journal of Computer Applications

摘　　要：成像误差修正是摄像机标定中的一个重要的环节,而传统的像差修正方法计算复杂。提出了一种基于BP神经网络的修正成像误差的算法,建立两个神经网络分别对δu与δv进行最佳函数逼近。仿真实验结果证明利用神经网络的非线性逼近能力无须建立精确的数学模型即可提高精度,而且比传统的方法更加简单易行。Correction of camera distortion is a main part of camera calibration. An algorithm of correction of camera distortion based on BP neural networks was proposed in this paper. Two neural networks were built to approximate the function of camera distortion. This algorithm can be considered as a modelfree solution to the problem of correction of camera distortion. The experimental results show that not only high accuracy can be obtained, but also the method is more simple compared with traditional methods.

关 键 词：BP网络 成像误差修正 逼近 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]