 |
 |
||
 |
|
||
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
||||||||
 |
Space Topics: Mars Reconnaissance OrbiterHiRISE Image Processing
Obtaining images from Mars Reconnaissance Orbiter isn't as simple as emailing a photo. There is an astonishing quantity of data in each HiRISE image. That data must be assembled, formatted, calibrated, stitched, mosaicked, and finally projected from the orbiter's point of view to a standard map orientation. This work is caried out at the HiRISE Operations Center (HiROC) in Tucson, Arizona, by a large team of scientists and students. This explanation of how HiRISE image processing works is based upon an entry in the HiRISE team blog by Ben Pearson, an undergraduate student who works on the technical support group for HiRISE. First, a note on the geometry of HiRISE images. HiRISE images are captured
in long swaths, parallel to MRO's roughly north-south track across Mars. HiRISE's
grayscale images are up to 20,264 pixels wide across each swath. To
capture such a large image, light from the optics is sent to an array of CCDs
that are each 2,048 pixels wide; 10 CCDs are required to cover the full swath
width, and there is a few pixels' worth of overlap between the CCDs. These
10 CCDs are sensitive to light in the red part of the visible light spectrum,
from 500 to 850 nanometers. Not all HiRISE images use the full 10 CCDs;
for some images, a subset of the CCDs are used, creating a narrower image. In addition, light from the center part of the HiRISE swath is sent to four more CCDs, two each that are sensitive to blue-green light (400 to 600 nanometers) and near-infrared light (800 to 1000 nanometers). Stacking views covering the same area in three different wavelengths allows HiRISE to obtain color images that are 4,048 pixels wide across the center of the swath. 1. Getting the data from Mars to EarthWhen HiRISE captures an image, it stores the data in up to 28 channels, two for each of the 14 CCDs. Each channel covers about half of the image. The image is placed inside a buffer on MRO, awaiting transmission to Earth, along with science data from the other instruments. At a scheduled time, a Deep Space Network (DSN) antenna in Spain, California,
or Australia -- usually one of the 34-meter antennas, more rarely the gigantic
70-meter antenna -- turns toward Mars, and MRO begins to downlink the data
from its buffer. The data is downlinked in packets and is transmitted
to the Jet Propulsion Laboratory (JPL) in Pasadena, California.
After each 4 hours of downlink, JPL puts the packets together for what is known as a "quick look." The quick look may only cover a portion of the full HiRISE image. JPL places this data on a server where it may be retrieved by computers at the HiRISE Operations Center (HiROC) at the University of Arizona's Lunar and Planetary Laboratory in Tucson, Arizona. A computer at HiROC periodically queries the servers at JPL to look for new data. When it finds new data, it copies the data to a server at HiROC, and image processing can begin, even if the image has not been completely received.
2. Assembling the dataFirst the image data is converted to a viewable format. These raw, unprocessed images are called Experiment Data Records, or EDRs. (This terminology is used for images and other types of data across all NASA and most ESA interplanetary missions.) Up to 28 EDRs are generated for each HiRISE image -- one EDR for each of the 28 possible channels of data. An EDR represents the data exactly as the HiRISE instrument captured it, and will be saved for archival purposes in NASA's Planetary Data System. In addition to image data, the EDRs include ancillary information that MRO returned to Earth about the image -- such as when the image was captured, how much it was compressed, and so forth. Next the images are converted to a format that is compatible with a software package called ISIS. ISIS, the Integrated Software for Imagers and Spectrometers, is a suite of tools developed by the United States Geological Survey to process images acquired from space missions. Most of the software tools used at HiROC were written to process images in ISIS format. Next the ISIS files -- still one for each of the 28 channels -- are calibrated. All
digital cameras contain some imperfections, such as irregularities in the
optics, "hot" pixels on the CCDs, and so forth. The calibration
step corrects the images for noise and these slight imperfections in the camera
system. An important goal of MRO's mission is to gather data that will
improve the calibration of the HiRISE images. For example, they capture
images of the night side of Mars, which should appear uniformly dark; variations
in the brightness of pixels in such a dark image result from variations in
the sensitivity of the individual detectors on the CCD. The nightside
image can be employed to correct daylit images for these sensitivity variations,
improving the quality of the data. Such measurements will be repeated
periodically throughout the mission, because the camera's performance will
degrade slightly over time.
Next, the calibrated files are run through a program called HiStitch, which joins together the pairs of images from each CCD. HiStitch generates up to 14 files, one for each CCD. Then a program called HiccdStitch assembles the images from each CCD into a mosaic representing the full swath. The CCDs acquire images with slightly overlapping coverage, so this step requires some processing. HiccdStitch generates up to three files, one for each wavelength band. All of these steps can be performed as soon as even a portion of the image data has been returned to Earth. At this point, however, if the image has not been completely received on Earth, processing halts. Occasionally, there are hiccups in the data transmission pipeline that can cause some gaps in images where data has been lost or corrupted. Sometimes these gaps can be filled with a second attempt to transmit the data from the spacecraft, but retransmission may not be higher priority than transmission of new data, or the spacecraft may need to erase its buffer in order to create room for more images, so some gaps are unrecoverable. Image processing continues once all data gaps have either been filled or have been determined to be unrecoverable. For images that included data from all three wavelength bands, the three bands can be assembled into one color image, a step that requires significant calibration and processing. 3. Getting a good look at the dataAll of the preceding steps are almost entirely automated; now, human eyes enter the pipeline. A team of students known as the HiRISE Validators examine the reprojected images to make sure that there have been no errors in the process. If they notice errors, they contact the HiRISE Operators, who attempt to resolve the problem, often passing part or all of an image through the processing pipeline again. In order to use the images for scientific research, they must first be reprojected to a standard map projection. HiRISE captures flat images from a tilted spacecraft above a lumpy, ellipsoidal planet. All flat images show a distorted perspective. Reprojection warps the image from the geometry of HiRISE's point of view to a map geometry with well-understood mathematical properties. This step allows scientists to measure distances and angles in HiRISE images to determine accurate sizes and shapes of features in the image and to compare them accurately to features in other images.
Geometric projection requires highly precise and accurate information about the position of the spacecraft with respect to Mars. This information is collected by MRO's navigational team when they perform Doppler tracking of the spacecraft, and is published in a data format known as a SPICE kernel (click here for more than you want to know about SPICE kernels). There is a time lag of up to two weeks between the acquisition of an image and the publication of the SPICE kernel that describes the position of MRO when the image is taken, so this step imposes a delay in image processing. It is possible to reproject an image based upon the predicted position of the spacecraft, and HiROC will do that for certain high-priority data sets. However, for most images, HiROC will wait for the correct SPICE kernels. The validated and reprojected images are converted from ISIS format to one that is more accessible to the public (and even to the science team), such as JPG or TIFF. HiRISE science team members examine the images to see if they contain anything particularly noteworthy. If so, they compose a caption for the image. In some cases they compose a press release, coordinating with the mission management at JPL to notify the world media about the noteworthy image. All of the images, captioned or not, are posted to the HiRISE and MRO websites, along with some of the image's ancillary information.
From start to finish, this process can take as little as a few hours to complete when an image can provide timely assistance to other Mars missions. This Victoria Crater image, for instance, was completed in 36 hours in order to get the data to the Opportunity operational team in time for them to plan the rover's next move toward the rim of the crater. However, for most images, the process will take approximately two weeks to complete. Getting all of these images out to the public is a high priority at HiROC, and they work hard to do it as fast as possible. This explanation of how HiRISE image processing works is based upon an entry in the HiRISE team blog by Ben Pearson, an undergraduate student who works on the technical support group for HiRISE. |
||||||||||||||
 |
 |
||||||||||||||
[ what is this? ]