| Author Name | Affiliation | | Lei Wan | Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Science, Changchun 130033, China
Graduate School of the Chinese Academy of Science, Beijing 100084, China
Key Laboratory of Airborne Optical Imaging and Measurement, Chinese Academy of Science, Changchun 130033, China | | Ye Zhang | Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Science, Changchun 130033, China
Key Laboratory of Airborne Optical Imaging and Measurement, Chinese Academy of Science, Changchun 130033, China | | Ping Jia | Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Science, Changchun 130033, China
Key Laboratory of Airborne Optical Imaging and Measurement, Chinese Academy of Science, Changchun 130033, China | | Jiajia Xu | Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Science, Changchun 130033, China
Key Laboratory of Airborne Optical Imaging and Measurement, Chinese Academy of Science, Changchun 130033, China |
|
| Abstract: |
| Due to the electronic rolling shutter, high-speed Complementary Metal-Oxide Semiconductor (CMOS) aerial cameras are generally subject to geometric distortions, which cannot be perfectly corrected by conventional vision-based algorithms. In this paper we propose a novel approach to address the problem of rolling shutter distortion in aerial imaging. A mathematical model is established by the coordinate transformation method. It can directly calculate the pixel distortion when an aerial camera is imaging at arbitrary gesture angles. Then all pixel distortions form a distortion map over the whole CMOS array and the map is exploited in the image rectification process incorporating reverse projection. The error analysis indicates that within the margin of measuring errors, the final calculation error of our model is less than 1/2 pixel. The experimental results show that our approach yields good rectification performance in a series of images with different distortions. We demonstrate that our method outperforms other vision-based algorithms in terms of the computational complexity, which makes it more suitable for aerial real-time imaging. |
| Key words: aerial camera CMOS sensor rolling shutter effect coordinate transformation image rectification |
| DOI：10.11916/j.issn.1005-9113.15342 |
| Clc Number:V248.1 |
| Fund: |
|
| Descriptions in Chinese: |
| CMOS航空相机卷帘快门效应的建模与校正 万磊1,2,3，张叶1,3，贾平1,3，许佳佳1,3 （1.中国科学院 长春光学精密机械与物理研究所，长春130033； 2.中国科学院大学研究生院，北京100084； 3.中国科学院 航空光学成像与测量重点实验室，长春130033） 研究目的：由于采用电子卷帘式快门，CMOS航空相机在高速成像时会产生几何畸变，这种畸变无法通过传统的基于视觉的方法完美校正。为了解决航空成像中的卷帘畸变问题，本文提出了一种基于任意姿态角航空成像模型的单帧畸变图像恢复方法。 研究方法： 首先通过坐标变换的方法建立了任意姿态角下CMOS航空相机成像模型，利用此模型计算相机在高速成像时像面上任意像素的畸变，由此组成覆盖整个CMOS像面的畸变地图。最后结合反向投影法对畸变图像进行校正。 结果： 1. 通过误差分析表明，在测量误差范围内，所建模型的最终计算误差小于1/2个像素； 2. 文中采集了一系列不同畸变类型和畸变大小的图像，分别进行了校正，结果证明了本文方法的有效性； 3. 文中将该方法和多种基于视觉的畸变校正方法在时间复杂度上进行了对比，结果表明本文方法比基于视觉的方法效率提高了10倍以上。 结论： 以上实验结果证明了本文方法的有效性，并且该方法的理论校正精度可以达到亚像素级。时间复杂度的对比分析表明相比基于视觉的校正方法，本文方法时间复杂度较低，更加适用于航空实时成像。 关键词：航空相机；CMOS传感器；卷帘快门效应；坐标变换；图像校正 |
