HiJACK: Correcting spacecraft jitter in HiRISE images of Mars
暂无分享,去创建一个
Since arriving at Mars orbit in March 2006, the High Resolution Imaging Science Experiment (HiRISE) onboard the Mars Reconnaissance Orbiter (MRO) has returned thousands of images of the surface of Mars at ~0.25 m/pixel ground sample distance [1]. The high resolution of HiRISE also makes it sensitive to geometric distortions due to spacecraft motion, which we call “jitter.” We describe ongoing work involved in modelling the jitter motion in each observation and implementing a solution to update the camera pointing to produce images with minimal geometric distortions. This processing pipeline will be called HiRISE Jitter-Analyzed CK, or HiJACK. Jitter correction as shown here improves results from digital elevation modelling using HiRISE stereo pairs [2]. The HiRISE operations team plans to release precision geometric products to the Planetary Data System (PDS), including updated pointing kernels for use by the wider scientific community. HiRISE Focal Plane Characteristics HiRISE is a pushbroom camera, with 14 ChargeCoupled Devices (CCDs) arranged in a staggered array on a fixed focal plane (Fig. 1). Each CCD is 2048 pixels wide. The focal plane is covered by a three-band filter, collecting red visible wavelengths (RED) across the full width of an image, with two additional detectors in the visible bluegreen (BG) and two in the near infrared (IR) in the center of the array. We take advantage of the cross-track overlap and the along-track time separation of the CCDs to gather information about the spacecraft motion [1]. Figure 1: HiRISE focal plane layout (not to scale). Resolving the Jitter We begin with a set of CCDs that overlap a common CCD (e.g. RED3-4, RED4-5, BG12RED4). Data from each CCD are radiometrically and geometrically calibrated [2, 3]. Any binned data are enlarged so the pixel dimensions from all CCDs are the same. We then measure the pixel coordinates of features in the overlap area of each CCD pair by using the ISIS 3 program hijitreg [4]. Relating the Offsets to Jitter The key to reconstructing the jitter in a HiRISE observation is the fixed and known along-track separation of each CCD pair. If the ground coordinates of the features were known a priori, we could solve directly for the jitter displacement j(t) between the expected and measured pixel coordinates of the feature at the time t of observation. Because the ground coordinates are not known, however, we can only measure the ‘offset’ f(t) between the feature position in the second CCD of the pair and the location we would predict based on the first CCD and steady jitterfree motion. This is described by the equation f (t) = j (t + t) j (t) (1) where t is the along-track time separation between the overlapping detectors. We estimate jitter separately in the x and y directions. After filtering to remove outlier points, the offset data are spline-interpolated to a uniform time sampling. The uniformly sampled offsets are then transformed into the frequency domain by a fast Fourier transform (FFT). In this domain, the difference equation (1) corresponds to an explicit algebraic relation between the Fourier series coefficients for offsets and those for jitter. The jitter coefficients are solved for, results for the three CCD pairs are combined as described below, and a jitter time series is obtained by inverse FFT. As a check, offsets are computed from the reconstructed jitter series and compared to the observations. Average error is typically 0.l–0.3 pixel (Fig. 2). Each CCD pair is effectively “blind” to jitter at frequencies with an integer number of cycles during the time offset t . HiRISE was designed to have different t for different CCD pairs, so that these blind spots fortunately do not overlap. Jitter coefficients at frequencies to which a given detector pair is blind are based on the average coefficients obtained from the other two pairs,