direct visual interpretation without dichotomous keys or in
components of the machine approaches such as signature
extraction or labeling of unsupervised signatures. Jensen
(2007) refers to this as the ‘art and science of remote sens-
ing’. Philipson (1980) provided one of the first attempts to
compare the methods of visual versus machine analysis in
a review of problem solving in remote sensing. Among his
observations was what has become almost the first rule of
remote sensing in that effective information extraction re-
quires about equal understanding of remote sensing and the
discipline or application of interest. Philipson established
four questions for problem definition and response:
1.
What are the targets (i.e. objects, keys, indicators)?
2.
What level of interpretation is necessary?
3.
How can the targets be sensed?
4.
What are the available resources and required data?
Question 2 separated feature identification (there is a build-
ing there) from feature interpretation (what the building is
used for) and employed the distinction between land cover
(e.g., building) from land use (e.g., residence). Question 3
was directed at the different, typically four, types of resolu-
tion in remote sensing, including spatial, spectral, tempo-
ral, and radiometric. After obtaining answers to these four
questions, Philipson (1980) suggests that the appropriate
information extraction method can be resolved based upon
the following five points:
1.
The need for visual interpretation increases with the
complexity of the problem.
2.
Computer methods are less useful for subsurface specu-
lations.
3.
As the required spatial and spectral sensitivity ap-
proaches the limits of the sensor, the need for using the
original remotely sensed data increases.
4.
Remote sensing cannot produce results of a specified re-
liability and geometric accuracy unless all of the remote
sensing data meet those levels.
5.
The costs of data and data analysis should be kept to a
minimum.
Philipson (1986) revisited the 1980 article to acknowledge
the increased resolutions of remote sensing data but also
the increased costs of the data. In essence, he observed that
improved data by themselves do not make visual or machine
the better approach as operationally, the overall project
costs are a significant factor. Ryerson (1989) also addressed
the role of traditional image interpretation with the advent
of spaceborne platforms and multispectral scanners and his
concern that ‘there has been too little attention paid to rig-
orous, visually-based image interpretation’. Ryerson (1989)
concluded with a look to the future where among other
observations he noted the increasing role of Geographic In-
formation Systems (GIS) as an asset in image interpretation
and that visual interpretation will remain a central element
in the information extraction process as many of the ele-
ments of image interpretation will be difficult to quantify.
T
extbooks
While there were early materials produced to assist in
the process of aerial photointerpretation, they were often
elaborations on specific photointerpretation keys, but they
did tend to focus on the understanding of what the photo-
graph portrayed and the extraction of information (Colwell
1954). After WWII there was an increase in university and
industrial or professional organizational aerial photographic
interpretation courses and a gradual increase in training
materials. Bianchetti and MacEachren (2015) examined
cognitive themes from early textbooks on the topic published
between 1922 and 1960. They evaluated 16 documents
including both governmental publications and educational
textbooks. A number of these were specific to military ac-
tivities while others were related to surveying and forestry.
They concluded that in these early materials the relation-
ships between perceptual cues and features were frequent
but that reasoning skill and knowledge were less common.
The ASP 1960 ‘Manual of Photointerpretation’ was one of
the truly significant contributions to early educational mate-
rials (Colwell 1960). It was followed by the Manual of Color
Aerial Photography (Smith and Anson 1968). The second
edition of the ‘Manual of Photointerpretation’ (Philipson
1997) is the standard reference book on this topic currently
available. There were also early textbooks among which was
‘Interpretation of Aerial Photographs’ by Avery and Berlin
(1962). This text was last published as the fifth edition in
1992 under the broader title of ‘Fundamentals of Remote
Sensing and Airphoto Interpretation,’ acknowledging the
changes in sensors and platforms.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
December 2017
797
