面阵相机几何畸变矫正方法研究  

The Research of Plane Array Camera Geometric Distortion Correction Method

作　　者：金元[1] 孟金宝[1] 王琨[1] 蔡汉辉[1] 席英[1] 朱近[1] 

机构地区：[1]南京理工大学计算机科学与技术学院,江苏南京210094

出　　处：《微电子学与计算机》2011年第10期36-38,43,共4页Microelectronics & Computer

摘　　要：由于机械加工安装、镜头光学特性等误差因素会使相机拍摄的图像产生非线性畸变,所以在应用所拍摄图像之前需要对图像进行畸变校正.在从硬件和软件两个方面对图像进行畸变校正的方法进行总结的基础上,具体介绍了基于BP神经网络的图像几何畸变矫正方法和网格靶标交点坐标的提取方法.实验结果表明,BP神经网络的图像几何畸变矫正方法简单易于实现,且能较全面地拟合畸变函数,得到高精度的校正图像.Because of the error factors such as machining installation and optical properties of the lens,The images will produce nonlinear distortion.The distorted images must be corrected before they are used.This paper presents BP neural networks method which is based on summarizing hardware and software methods of correction and intersection extraction in details.Experimental results show that both are easy to implement,and BP neural network method can fit the distortion function of a more comprehensive,and obtain high-precision calibration images.

关 键 词：畸变矫正 交点提取 BP神经网络 图像变形 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]