面阵CCD航空相机斜视图像的几何校正

为增大机载光电侦察系统的视场角,航空相机多采用扫描方式对地摄影成像,文章针对面阵CCD航空相机斜视摄影引起的图像变形进行校正。由于该类相机与载机之间存在方位和俯仰两轴运动,文章同时考虑相机相对载机的旋转角度和载机相对地面的姿态角,建立了更符合该类相机工作状态的六姿态角投影校正模型,推导出同一地物在畸变图像和标准图像上的像素坐标变换关系,采用双线性插值算法对坐标变换后的像素灰度值进行重采样,对航空图像进行自动校正。最后,对某机场的航拍图像进行校正实验,并与基于地面控制点的多项式校正方法以及基于畸变图像和参考图像配准的校正方法进行比较。实验结果表明,六姿态角投影校正模型能够取得比较高的校正精度(可达像素级),当姿态角的测量精度为3′,图像大小为512pixel×512pixel时,图像校正的均方根误差为1.174 7pixel。该方法不需要提供参考图像和野外采集地面控制点数据,便于工程实现,基本满足航空图像处理的稳定可靠、精度高、实时性强等要求。

Geometric correction of oblique images for array CCD aerial cameras

To increase the airborne photoelectric reconnaissance devices' field of view,most aerial cameras adopt the method of scanning to photograph the earth. For the purpose of image distortion caused by the oblique photography of array CCD aerial camera, a practical auto-rectification approach was proposed in the thesis.As there was a dual-axis rotation (specifically,the azimuth and pitch motion) between these array CCD aerial cameras and aircrafts, this paper introduced a new approach for image geometric correction that took into account both the rotational angles from cameras to aircrafts and the attitude angles of the aircrafts relative to the ground. First, the six attitude angles projective rectification model of these array CCD aerial cameras was established. Then, the relationship of pixel coordinates between distortion image and standard image was obtained. Through coordinate transformation and luminance bilinear interpolation, the aerial remote sensing image with distortion was auto-accurately rectificated. Finally, the aerial image of an airport was corrected using the six attitude angles projective rectification model, the ground control point based multinomial correction model and the correction method of matching the distortion image and reference image, respectively. Experiments show that the six attitude angles projective rectification model can achieve pixel level correction precision. For 512 pixel×512 pixel images,the root mean square error of image correction can reach 1.174 7 pixel accuracy when the accuracy of attitude angles is 3 min.The proposed algorithm in the thesis doesn't need for reference images and ground control points, and facilitates engineering application. It can satisfy the system requirements of online, real time, higher precision and stabilization.