High Resolution Space Camera Image Motion Measurement Based On Images From TDICCD Overlapping Area

Spacecraft orbital motion, Earth rotation, spacecraft attitude change and vibrations of moving parts will degrade images by causing image motion on focal plane in space camera, image motion compensation is one of the most effective means to weakening the influence, image motion measurement is the essential prerequisite for motion compensation. This paper focuses on the image motion estimation method for high resolution TDICCD visible light Earth observation camera, which not only is of great significance for weakening the image degradation caused by image motion, but also contributes to image distortion correction and knowledge for spacecraft vibration.Given that the imaging position may be different in overlapping images from two adjacent TDICCDs due to the image motion, this paper proposes an image motion estimation method using the image position differences. This approach has three advantages. First, it tries to measure the image motion at the most important location of a space camera: the focal plane, where the imaging sensor is located, the estimated image motion is much more close to the true value at the specific imaging moment. Second, it does have a wide measurement band of image motion. Third, it’s less susceptible to vibration disturbance.This paper describes the principle and implementation of the image motion estimation method, analyzes the image motion measurement error, time consuming and bandwidth. The relationship between measurement error and image content, image blurring and signal-to-noise is analyzed, the MTF of a space camera and the image SNR have to be more than 0.055 and 22 to ensure the image motion measurement error not more than 1 pixel in the use of phase correlation sub-pixel image registration method. The hardware implementation time of image registration is the key factor affecting image motion measurement bandwidth, the image motion measurement implementation bandwidth can achieve 200 Hz under the current level of technology.Image motion initial values are key for the subsequent image motion measurement, in order to obtain the initials, firstly, this paper establishes a low-frequency image motion time-domain model and takes it as the fitting function, the low-frequency image motion results from the Earth observation model are taken as observation points, then the fitting curve can be obtained by Fourier transformation of the observation points. Secondly, a minimizing criterion function is established, and the minimum of which is taken as the image motion estimation initial values. Finally, all factors influencing image motion initials estimation and their transfer coefficient are analyzed, the results shows that offset error is the main factor. In addition, the initials error also causes steps in image motion curve, smoothing is used to remove the steps.The existence of blind band is analyzed and the blind bands locations are determined by the number of lines in space between two overlapping CCDs and their line time. Firstly, this paper proposes an improved layout of focal plane in order to eliminate blind band in image motion estimation, on one hand, the two numbers of lines in space of two overlapping CCD pairs must be coprime in order to minimize the aliasing range of blind band; on the other hand, the paper proposes indirect image motion estimation method using a non-overlapping CCDs pair with much smaller space for measuring the image motion at aliasing frequencies within bandwidth. Secondly, the influence of the error of lines in space and the line time variation on the blind band elimination are analyzed, this paper also gives the spans of the bandwidth of image motion measurement and related parameters of focus plane for ignoring all error. Finally, the paper gives the perfect location of the 3 CCDs used for image motion measurement under rolling imaging.Two separated experiments are carried out to prove the correctness of the image motion measurement method, the results show that this method can effectively achieve wide-frequency-range measurement of image motion.