基于地理位置信息的图像配准方法及精度分析

Image Registration Method Based on Geo-Location Information and Precision Analysis

针对传统图像配准技术难以对海洋、沙漠、草原等特征不明显区域航空遥感图像进行配准的问题,提出了一种基于地理位置信息的航空遥感图像配准算法。依据载机定位定向系统测量的载机位置、姿态信息以及航空相机中位置编码器测量的框架角位置信息,利用齐次坐标变换的方法求解配准点在大地坐标系下的投影。利用世界大地坐标系-84坐标系定义的地球椭球模型确定匹配点的经纬度信息,将相同地理位置信息的配准点进行配准。采用蒙特卡罗法仿真分析了载机姿态位置信息及框架角位置信息对配准精度及定位精度的影响。采用实际的航空遥感图像进行实验,结果表明,在载机飞行高度低于2000m,拍摄倾斜角小于18°时,配准精度可优于3m,遥感图像中的海上控制点的定位精度优于35m。

Traditional image registration techniques are inadaptable to aerial remote sensing image which does not have obvious feature points,such as ocean,desert,and grassland.A registration algorithm is proposed for aerial remote sensing image based on geo-location information.Using aircraft position and attitude information measured by airborne position and orientation system and the gimbal angle from the encoder in an aerial camera,the projection of registration point in geodetic coordinate system is solved by homogeneous coordinate transformation.Using the earth ellipsoidal model which is defined by the world geodetic system-84,the longitude and latitude information for matching points are solved,and the matching points with the same geo-location information are registered.The influence of the aircraft position and attitude and the gimbal angle on geo-location accuracy and registration accuracy is analyzed with Monte Carlo method.Experiments were carried out using actual aerial remote sensing images.Experimental results show that the registration accuracy is less than 3m,and the positioning accuracy of sea control point in remote sensing image is less than 35 m,when the flight height of plane is below 2000 m and the photography inclination angle is less than 18°.