Data Set Overview
    =================
      The Clementine Basemap Mosaic is a full-resolution (100 meters
      per pixel) global mosaic produced by the U.S. Geological Survey
      from Clementine EDR Data. The 750 nanometer filter imaging
      data acquired by the Ultraviolet/Visible Camera were used to
      create the single-band basemap mosaic.
 
      The Clementine Basemap Mosaic is partitioned on the CD
      collection in the Sinusoidal Equal-Area Projection as 12
      'zones', each 30 degrees wide in longitude and ranging from 70
      degrees south latitude to 70 degrees north latitude. All tiles
      in a zone have the same center longitude of projection. Both
      polar regions between 70 degrees to 90 degrees latitude exist
      both as  Sinusoidal and polar stereographic projections. Each
      zone and each polar region exist on one CD-ROM volume. This
      results in a 14-volume archive set containing the full
      resolution (0.1 km/pixel) mosaic. A fifteenth volume,
      containing reduced-resolution planetwide coverage at .5, 2.5,
      and 12.5 km/pixel and other ancillary data, complete the
      archive collection.  For each full- and reduced- resolution
      image product, a sub-sampled 'browse' image is provided in
      Joint Photographic Experts Group (JPEG) format.
 
      Each 30 degree zone is further divided into smaller tiles. The
      tiling scheme, basic to digital cartography design, is similar
      to previous planetary global mosaics, and maintains reasonably
      sized image products. In general, this design consists of
      rectangular tiles that are  roughly 2100 pixels on a side. The
      actual tile size varies with latitude. Near the equator, each
      tile covers 7 degrees of latitude and 6 degrees of longitude. A
      typical full-resolution tile required ~9 megabytes of digital
      storage and may  contain approximately 35 raw Clementine
      images.
 
 
    Parameters
    ==========
      N/A
 
 
    Processing
    ==========
      The Integrated Software for Imaging Spectrometers (ISIS)
      processing system, developed by the U.S. Geological Survey
      was used to generate the basemap mosaic.  Processing within
      ISIS includes radiometric and geometric correction, spectral
      registration, photometric normalization, and image mosaicking.
      Radiometric correction applies 'flat fielding', dark current
      subtraction, non-linearity correction, and conversion to
      radiometric units. Geometric transformations tie each raw image
      with a ground control network and convert from raw image
      coordinates to the Sinusoidal Equal-Area projection.
      Photometric normalization is applied to balance brightness
      variations due to illumination differences among the images
      in a mosaic. Images are then mosaicked together to form a
      global map of continuous image coverage for the entire planet.
 
 
    Media/Format
    ============
      The Clementine basemap is delivered to the Planetary Data
      System using CD media.  Formats are based on standards for
      such products established by the Planetary Data System (PDS)
      [PDSSR1992].

Overview
      ========
      The Clementine Basemap Mosaic is the result of an
      exhaustive Lunar cartography project based on data from
      the Clementine EDR image collection. Systematic calibration
      and processing enable global, full-resolution scientific
      analysis of the Clementine Datasets. A first major step
      in the systematic processing of the imaging data is the
      production of an accurate basemap to which all products
      are geometrically registered. Previous maps and ground
      control points of he Moon is not sufficiently accurate
      The previous RAND control network is accurate to 500
      meters in the area covered by the Apollo mapping frames
      (15% of the Moon's surface), and is accurate to about
      1-2 kilometers for regions covered by telescopic,
      Galileo, and Mariner 10 observations. However, most
      of the far side is not included in the network, and the
      only other positional dataset for these regions contains
      errors as large as tens of kilometers. Based on best
      effort measurements of the spacecraft orbit and pointing,
      UVVIS geometric distortions, and time tags for each
      observation, the SPICE data alone provides positional
      accuracies better than 1 kilometer over most of the
      Moon. With residuals primarily small random pointing
      errors, then accuracies approaching the UVVIS scale
      becomes achievable.
 
      The goal of the basemap is for 99% of the Moon
      (excluding the oblique observation gap fills) to
      be better than 0.5 km/pixel absolute positional accuracy
      and to adjust the camera angles so that all frames
      match neighboring frames to within an accuracy of 2 pixels.
      To achieve these goals we required camera alignment and
      pointing data accurate to a few hundredths of a degree.
      We determined the absolute alignment of the UVVIS
      with respect to spacecraft-fixed axes (A and B Star
      Tracker Camera quaternions) by analyzing a major subset
      of the over 17,000 images of Vega, over 6,000 images of
      the Southern Cross and a few hundred images of the
      Pleiades, taken during the approach to the Moon and
      throughout the lunar mapping mission phase. Multiple
      star images within a single picture were used to determine
      the UVVIS focal length and optical distortion parameter
      values.
 
 
      Approximately 265,000 match points were collected at the
      USGS from ~43,000 UVVIS images providing global coverage.
      About 80% of these points were collected via autonomous
      procedures, whereas the 20% required the more time consuming
      but highly accurate pattern-recognition capability of the
      human eye-brain. We also developed streamlined procedures
      for the supervised collection of match points. The new
      procedures saved several person-years of effort and represents
      new capabilities useful with other planetary datasets. The
      automated success rate exceeded 90% along each spacecraft
      orbit track, where the overlap regions of successive
      images are highly correlated, but failed when the overlap
      regions is narrow and/or nearly featureless. ('Failure' is
      defined as less than 3 points per image with correlation
      coefficients grater than 0.85; thus, many good match points
      were rejected because we could not be certain that the
      matches were valid without verification.) Across-track
      matching was more difficult due to changes in scale and
      illumination angle, but a fair success rate (~60%) was
      nevertheless achieved via the use of 'window-shaping'
      (local geometric reprojections). The oblique gap-fill images
      were the most difficult to match, and required substantial
      human intervention. Matching the polar regions was
      time-consuming because each frame overlaps many other frames.
      most match points were found to a precision of 0.2 pixels.
 
      The USGS match points were sent to RAND corporation for
      analytical triangulations. Using these match points,
      control points from the Apollo region, and the latest
      NAIF/SPICE information, RAND determined improved
      camera orientation angles for the global set of UVVIS
      images. A constant lunar radius of 1737.4 kilometers
      was assumed, a significant source of error near the
      oblique gap fills. The analytical triangulation is a
      least-squares formulation designed to adjust the latitude
      and longitude of the control points and the camera
      orientation angles to best fit the match points. The
      triangulation was first computed on 'packets' of match
      points (each covering ~1/8-th of the Moon), then checked
      and rechecked at the USGS via plots and test mosaics to
      fix and add match points as needed. The final (global)
      analytical triangulation required solving ~660,000 normal
      equations. The mean error is less than 1 pixel. This is
      by far the largest analytical triangulation ever applied
      to a planetary body other than Earth. The results fully
      define the planimetric geometry of the basemap, to which
      future systematic products will be tied.