Robinson, J. A., D. A. Liddle, C. A. Evans, and D. L. Amsbury. 2002. Astronaut-acquired orbital photographs as digital data for remote sensing: spatial resolution. International Journal of Remote Sensing, 23(20):4403-4438.

Astronaut-acquired Orbital Photographs as Digital Data for Remote Sensing: Spatial Resolution

Click here for the PDF version

Menu Introduction Background Factors that determine the footprint Other characteristics that influence spatial resolution Estimating spatial resolution of astronaut photographs Summary and Conclusions 3. Factors that determine the footprint (area covered by the photograph)

Figure 1     The most fundamental metric that forms the basis for estimating spatial resolution of astronaut photographs is the size of the footprint, or area on the ground captured in a photograph (see review of satellite photogrammetry by Light 1980). The basic geometric variables that influence the area covered by an astronaut photograph are (1) altitude of the orbit, H, (2) focal length of the lens, f, (3) actual size of the image on the film, d, and (4) the orientation of the camera axis relative to the ground (the obliquity or look angle, t). The relationships among these parameters are illustrated in figure 1.

3.1. Altitude
Figure 2      Human spaceflight missions have had a variety of primary objectives that required different orbital altitudes. The higher the altitude, the larger the footprint of the photographs. The differing scale of photographs taken at different altitudes is illustrated in figure 2. The two photographs of Lake Eyre, Australia were taken with the same camera and lens, but on different dates and from different altitudes. In A, the lake is relatively flooded, while in B it is dry. A 2.3 ´ difference in altitude leads to a corresponding difference in the scales of the resulting photographs.

Figure 33.2. Lenses
     The longer the lens focal length, the more magnification, the greater detail, and the smaller footprint. A variety of lenses with different focal lengths are flown on each space mission (table 1). The effect of lens length on spatial coverage and image detail is shown in figure 3. In these views of Houston (A-C), taken from approximately similar altitudes with the same camera, most of the difference in scale of the photographs is due to the different magnifications of the lenses. Although taken from a different altitude and using a different camera with a different original image size (see table 2), we also include an electronic still camera image taken with a 300 mm lens for comparison.

Table1     For remote sensing, longer focal length lenses are generally preferred (250- or 350-mm for Hasselblad; 300- or 400-mm lenses for 35-mm format cameras and ESCs). Unfortunately, longer-focal-length lenses exhibit poorer performance toward the edge of a frame. For example, a 250 mm lens (Distagon CF 5.6) used on the Hasselblad camera has spatial resolution of 57 lp/mm at the centre, but only 51 lp/mm at the edge and 46 lp/mm at the corner (tested with Ektachrome 5017, f/8 aperture and high-contrast). A longer 300 mm lens used on the Nikon camera has a greater spatial resolution difference between centre (82 lp/mm) and corner (49 lp/mm, f/4 aperture, Kodak 5017 Ektachrome, high contrast, Fred Pearce, unpubl. data). There are tradeoffs among lens optics and speed. For example, the lenses for the Linhof system, and the 250-mm lens for the Hasselblad (Distagon CF 5.6, see footnotes to table 1) are limited to apertures smaller than f/5.6. This becomes an important constraint in selecting shutter speeds (see discussion of shutter speed in section 4.2, below).

3.3. Cameras and actual image sizes
Table2      After passing through the lens, the photographic image is projected onto film inside the camera. The size of this original image is another important property determining spatial resolution, and is determined by the camera used. Camera formats include 35-mm and 70-mm formats (Lowman 1980, Amsbury 1989); and occasionally 4 ´ 5-inch (101 ´ 127 mm) and larger formats (table 1). . Cameras flown on each mission are not metric—they lack vacuum platens or reseau grids, image-motion compensation, or gyro-stabilised mounts. The workhorse for engineering and Earth photography on NASA missions has been a series of 70-mm Hasselblad cameras (table 1), chosen for their reliability. The modified magazine databack imprints a unique number and timestamp on each frame at the time of exposure. Cameras are serviced between flights. Occasionally there is enough volume and mass allowance so that a Linhof 4 ´ 5-inch (101 ´ 127 mm) format camera can be flown. Nikon 35-mm cameras are flown routinely, also because of proven reliability. Electronic still cameras (ESC) were tested for Earth photography beginning in 1992 (Lulla and Holland 1993). In an ESC, a CCD (charge-coupled device) is used as a digital replacement for film recording the image projected inside the camera. An ESC (consisting of a Kodak DCS 460c CCD in a Nikon N-90S body) was added as routine equipment for handheld photographs in 1995. ESCs have also been operated remotely to capture and downlink Earth images through a NASA-sponsored educational program (EarthKAM). Discussion of CCD spatial array and radiometric sensitivity are beyond the scope of this paper, but are summarised by Robinson et al. (2000b). The format of the film (or CCD), and image size projected onto the film (or CCD) are summarised for all the different cameras flown in table 2.

3.4. Look angle or obliquity
Figure 4     No handheld photographs can be considered perfectly nadir; they are taken at a variety of look angles ranging from near vertical (looking down at approximately the nadir position of the spacecraft) to high oblique (images that include the curvature of Earth). Imaging at oblique look angles leads to an image where scale degrades away from nadir. A set of views of the same area form different look angles is shown in figure 4. The first two shots of the island of Hawaii were taken only a few seconds apart, and with the same lens. The third photograph was taken on a subsequent orbit and with a shorter lens. The curvature of the Earth can be seen in the upper left corner.

Figure 5     Obliquity can be described qualitatively (figure 4) or quantitatively as the look angle (t, figure 1, calculations described in Formulation 2 [section 5.2]). Because obliquity and look angle have such a dramatic influence on the footprint, we summarise the database characteristics relative to these two parameters. Figure 5 is a breakdown of the spatial resolution characteristics of low oblique and near vertical photographs in the NASA Astronaut Photography database. Number of photographs are grouped (A) by calculated values for look angle (t) and (B) by altitude. After observing the overlap between near vertical and low oblique classes, we are currently restructuring this variable ('tilt') in the database to provide a measure of t when available. Users will still be able to do searches based on the qualitative measures, but these measures will be more closely tied to actual look angle.

3.4.1. Obliquity and georeferencing digitised photographs
     Often the first step in a remote sensing analysis of a digitised astronaut photograph is to georeference the data and resample it to conform to a known map projection. Details and recommendations for resampling astronaut photography data are provided by Robinson et al. (2000a, c) and a tutorial is also available (McRay et al. 2000). Slightly oblique photographs can be geometrically corrected for remote sensing purposes, but extremely oblique photographs are not suited for geometric correction. When obliquity is too great, the spatial scale far away from nadir is much larger than the spatial scale closer to nadir; resampling results are unsuitable because pixels near nadir are lost as the image is resampled while many pixels far away from nadir are excessively replicated by resampling.

     To avoid the generation of excess pixels during georeferencing, the pixel sizes of the original digitised image should be smaller than the pixels in the final resampled image. Calculations of original pixel size using methods presented below can be useful in insuring meaningful resampling. For slightly oblique images, formulation 3 (section 5.3) can be used to estimate pixel sizes at various locations in a photograph (near nadir and away from nadir), and these pixel sizes then used to determine a reasonable pixel scale following resampling.

<< Back            Next >>