observations (usually due to the limitations of manpower). Photography is
relatively inexpensive, and because of its on-demand nature, allows planning for
optimum conditions (e.g., sun angle, cloud cover, sea-surface state, tide level)
during image collection. Distortions present in photographs are primarily a function of camera properties (e.g., radial lens distortions), which can be corrected
through use of camera models and geometric algorithms accounting for variations
in aircraft positioning (e.g., altitude, roll, pitch and yaw). Photographs are commonly collected from a vertical viewpoint (i.e., directly overhead), but in some
instances are collected from an oblique angle, which introduces additional distortions to be considered during analysis but provides a different viewpoint that
can be useful in some applications.
Computer analysis of film-based photographs first requires that they are scanned, or digitized, at a specified resolution (i.e., pixel resolution) to assign a digital
number (DN) representing the relative intensity of reflected light for each pixel in
the image. Alternately, for digital-based photographs, DN values already represent
the relative light intensity for specific spectral bands. The spatial resolution of each
pixel is dependent on the resolution of the camera, the scanner, and the flying
altitude upon collection, but is typically high enough (ranging from 0.1 to 1 m per
pixel) to be able to resolve many complex reef features.
Aerial photographs are usually collected on-demand, and scheduled when both
sky and water conditions are at their best. Nonetheless, since film photography
lacks detailed radiometric information, it is infeasible to apply preprocessing
corrections to the water column to account for attenuation of wavelengths with
depth due to absorption and scattering. In some instances, such corrections can be
applied when using digital aerial photography, but are best suited for multispectral
and hyperspectral sensors, which are spectrally better characterized than photography. As an example, Lyzenga (1978, 1981) describes the use of the image itself
to produce a depth-invariant bottom index to compensate for differences in variable depths when discriminating between bottom types.
Registration, or georeferencing, of aerial photographs collected for coral reef
studies may be difficult without supplemental data. Aerial photographs of on-land
features are georegistered using ground control points, previous orthophotographs,
and/or digital terrain models (DTMs). However, due to the offshore location of
many coral reef areas, images of these study sites may not contain any referencing
land or shoreline that can be matched to image locations, or the images may only
contain a limited linear extent of coastline, which, if paired with on-land ground
control points or a DTM and stretched to match, may spatially distort the seaward
edge of an image. In these instances, it may be necessary to use georeferenced
bathymetric data, such as LiDAR (Light Detection and Ranging; see Chaps. 5–7)
or other underwater terrain models, to correctly georeference the aerial images.
Additionally, because astronaut-acquired photographs are acquired using varying
camera lenses from different altitudes, resulting in different scales, and are also
taken from different look angles through the windows of spacecraft, which leads to
different spatial resolutions between the near and far edges of a photograph,
additional steps must be taken to resample and geometrically correct these images
2 Photography Application
33
relatively inexpensive, and because of its on-demand nature, allows planning for
optimum conditions (e.g., sun angle, cloud cover, sea-surface state, tide level)
during image collection. Distortions present in photographs are primarily a function of camera properties (e.g., radial lens distortions), which can be corrected
through use of camera models and geometric algorithms accounting for variations
in aircraft positioning (e.g., altitude, roll, pitch and yaw). Photographs are commonly collected from a vertical viewpoint (i.e., directly overhead), but in some
instances are collected from an oblique angle, which introduces additional distortions to be considered during analysis but provides a different viewpoint that
can be useful in some applications.
Computer analysis of film-based photographs first requires that they are scanned, or digitized, at a specified resolution (i.e., pixel resolution) to assign a digital
number (DN) representing the relative intensity of reflected light for each pixel in
the image. Alternately, for digital-based photographs, DN values already represent
the relative light intensity for specific spectral bands. The spatial resolution of each
pixel is dependent on the resolution of the camera, the scanner, and the flying
altitude upon collection, but is typically high enough (ranging from 0.1 to 1 m per
pixel) to be able to resolve many complex reef features.
Aerial photographs are usually collected on-demand, and scheduled when both
sky and water conditions are at their best. Nonetheless, since film photography
lacks detailed radiometric information, it is infeasible to apply preprocessing
corrections to the water column to account for attenuation of wavelengths with
depth due to absorption and scattering. In some instances, such corrections can be
applied when using digital aerial photography, but are best suited for multispectral
and hyperspectral sensors, which are spectrally better characterized than photography. As an example, Lyzenga (1978, 1981) describes the use of the image itself
to produce a depth-invariant bottom index to compensate for differences in variable depths when discriminating between bottom types.
Registration, or georeferencing, of aerial photographs collected for coral reef
studies may be difficult without supplemental data. Aerial photographs of on-land
features are georegistered using ground control points, previous orthophotographs,
and/or digital terrain models (DTMs). However, due to the offshore location of
many coral reef areas, images of these study sites may not contain any referencing
land or shoreline that can be matched to image locations, or the images may only
contain a limited linear extent of coastline, which, if paired with on-land ground
control points or a DTM and stretched to match, may spatially distort the seaward
edge of an image. In these instances, it may be necessary to use georeferenced
bathymetric data, such as LiDAR (Light Detection and Ranging; see Chaps. 5–7)
or other underwater terrain models, to correctly georeference the aerial images.
Additionally, because astronaut-acquired photographs are acquired using varying
camera lenses from different altitudes, resulting in different scales, and are also
taken from different look angles through the windows of spacecraft, which leads to
different spatial resolutions between the near and far edges of a photograph,
additional steps must be taken to resample and geometrically correct these images
2 Photography Application
33