754
December 2018
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
BOOK
REVIEW
Landsat’s
Enduring
Legacy
and many
other ASPRS
publications
available at
Amazon.com.
LANDSAT’S ENDURING LEGACY
PioneeringGlobalLandObservations fromSpace
LandsatLegacyProjectTeam
THE
IMAGING & GEOSPATIAL
INFORMATION SOCIETY
Order online at amazon.com
report the results of applying his 3D scene analysis framework
to multiple benchmark point cloud data sets using multiple
configurations that give the reader a wider understanding of
what the technology is capable of accomplishing. Additional-
ly, beyond the level of simply reporting results from standard
procedures, Dr. Weinmann advances the state of the art by
answering what would appear to be fundamental questions
regarding the interactions affecting workflow among the pro-
cessing elements that comprise his framework (Figure 1).
y
y
Neighborhood Selection — ensures that the local descrip-
tion of a feature’s 3D geometric structure is contained
within the selected neighborhood by relying on the spatial
arrangement of all 3D points in the neighborhood
y
y
Feature Extraction — addresses the design of suitable
features
y
y
Optimized Feature Selection — addresses the Hughes
phenomenon, where an increasing number of features will
decrease a classifier’s predictive power
y
y
Supervised Classification — considers classifier selection
and performance in the context of optimizing all four el-
ements to satisfy constraints of simplicity, efficiency and
reproducibility
Reconstruction
succeeds at multiple levels, making it a
worthwhile addition to your reference shelf. As a research
monograph,
Reconstruction
is a fantastic resource regarding
work that has been done in point cloud processing. This book
also shines as a practical exploration and comparison of fun-
damental point cloud processing tasks and algorithms, which
will give researchers, developers and practitioners a basis for
choosing their own candidate path through a forest of point
cloud processing methods, or as a starting point for their own
research. As original work, Dr. Weinmann clearly identifies
the contributions he has made and the broader motivations for
addressing a problem in the first place.
From the perspective of a practitioner seeking to under-
stand the point cloud processing tools that he or she may be
working with,
Reconstruction
provides a great entrée into
developing an understanding of the existing body of work.
Although it was published in 2016, and this is a rapidly mov-
ing field, it is certainly timely enough that it should provide
a sufficient basis for getting up to speed with the most recent
developments. However, that is not always necessary, be-
cause for a book as recent as this one, the most cutting edge
research is often not implemented yet in off the shelf soft-
ware. What is usually more frustrating for the practitioner
is the black box nature of the software that one works with,
and trying to get an idea what is going on under the hood
when you are making a buying decision. Hopefully this book
can arm you with sufficient knowledge to ask your vendor
intelligent questions.
From the perspective of the researcher or developer, you
can compare your own bibliography with Dr. Weinmann’s
extensive citations to see where yours may be lacking, or
understand how his perspective might differ from your own.
Personally, thinking of reconstruction, as referenced in the
title, puts me more in mind of the 3D visualization and
simulation and the geometry of closed 3D shapes or objects.
Once it became apparent that Dr. Weinmann’s focus was on
the exploitation of an object’s local surface structure infor-
mation, or representing arbitrary 3D shapes,
3
the sources
cited looked appropriate, current and fairly comprehensive.
They included sources associated with computer vision and
photogrammetry & remote sensing. Finally, as noted above,
the structure of the book makes it very accessible as both a
reference work and a snapshot report of a researcher’s work,
and perhaps the best part from a research perspective is that
Dr. Weinmann continues his work in the field, publishes
extensively,
4
and makes software available.
5
In summary,
Reconstruction
is successful in achieving
multiple objectives and in potentially appealing to multiple
audiences. While not intended as a textbook due to the short
shelf life of the specific work in presents, its collection and
review of fundamentals and related work means it could very
well serve as a text for a focused graduate or undergradu-
ate course. It certainly deserves reading and a place on the
reference shelf for researchers, developers and practitioners.
All of the audiences would also be well served to track Dr.
Weinmann’s ongoing research to see where work in this field
is heading.
3
Reconstruction
, p 46.
4
5
Reconstruction
, p 213, and
