Detection of Satellite Attitude Jitter Based on Image Processing

Monday, 15 December 2014
Shijie Liu1, Xiaohua Tong1, Zhen Ye1, Xinming Tang2, Yusheng Xu1, Lingyun Li1, Fengxiang Wang1, Huan Xie1, Junfeng Xie2 and Tianpeng Li1, (1)Tongji University, Shanghai, China, (2)Satellite Surveying and Mapping Application Center, National Administration of Surveying, Mapping and Geoinformation, Beijing, China
High-resolution satellite imageries (HRSIs) always suffer from mechanical vibration during scan, resulting in attitude jitter and non-ignorable errors in geo-positioning and mapping. Therefore, it is critical to detect and estimate the attitude jitter for further possible compensation to explore the full geometric potential of HRSI. We bring up with a solution to detect the attitude jitter effect based on image processing using images recorded by a sensor system with parallax observation.

Three methods of attitude jitter detection are investigated. The first one is based on analysis of the co-registration errors between images with very small parallax observation (e.g. different bands of multispectral image). The second one is based on stereo images using sensor imaging models to investigate the geometric inconsistance in image space. The third one is based on analysis of the co-registration errors of two DOM products from the images. Phase correlation, geometric constraint cross correlation and least squares matching are used in our methods correspondingly for high accuracy image matching, and the RANSAC algorithm is adopted to remove mismatched points and outliers. Finally, the image disparities from each method are used to investigate the effect and characteristic of satellite attitude jitter.

We applied our methods on different satellites to investigate their attitude jitter characteristics. Results of experiment with multispectral images obtained by the ASTER camera equipped on Terra satellite showed that there exist more than one frequency with amplitude up to 0.3 pixel. Experimental results with panchromatic image strips captured by LROC revealed that there exist at least two attitude jitter frequencies with amplitude up to 0.6 pixel. Three methods were all used to investigate the attitude jitter of Chinese ZY-3 satellite and the results from different methods showed good consistency, and a distinct periodic attitude fluctuation with frequency around 0.65Hz always exist, and the amplitude decreased from the early 1 pixel down to within 0.5 pixel, revealing that ZY-3 platform becomes steady with good potential of high positioning accuracy.

Key words: satellite attitude jitter, high resolution satellite imagery, ASTER, LROC, ZY-3