had the most overall image artifacts with 89 percent of the
sites showing some degree of artifact (Figure 4). This was sig-
nificantly more than Pix4D or
ICE
with 29 percent and 0 per-
cent, respectively, (Chi Square p <0.001) (Table 5). Pix4D also
produced significantly more than
ICE
(Chi Square p <0.001). It
is likely that the cause for the lack of image artifacts produced
by
ICE
is due to the different approach its algorithms take
to create the mosaic;
ICE
maintains the original image pix-
els rather than blending pixels from multiple images at the
expense of geometric accuracy and occasionally consistency
between images.
When comparing the amount of image artifacts between
the software packages with respect to different vegetation
types on the ground, Photoscan Pro was found to produce sig-
nificantly more artifacts in every vegetation category (Fisher’s
exact p <0.001 for all comparisons) (Table 5). In fact, the only
comparison found not be significant was wet-mesic prairie be-
tween
ICE
and Pix4D (Fisher’s exact
p
= 0.306). The
p
value of
0.019 found for typha marsh between
ICE
and Pix4D, although
technically above the threshold of 0.017 was still considered
significant due the close proximity with regards to the target
value and the conservative nature of Bonferroni corrections
(McDonald, 2009).
Although differences were found between the different
software packages when comparing image artifact presence
with the vegetation types, no significant differences were
found when comparing the vegetation types within a single
software package (i.e., canary reed grass versus wet-mesic
prairie). This indicates that image artifacts, although more
likely to occur in Photoscan Pro, are not affected by differenc-
es in the structure of the vegetation such as leaf size, height,
or woody versus. herbaceous.
Pix4D showed the highest incident of overall image blur
with 38 percent of the total sites showing some level of blur
(Figure 5). This was found to significantly higher than Photo-
scan Pro with 14 percent and
ICE
with 20 percent (Chi Square
p <0.001 and p = 0.05 respectively) (Table 6). Photoscan Pro
and
ICE
were not found to significantly different (Chi Square p
= 0.259). It should be noted that there is some inherent blur in
the original imagery due to motion from the aircraft platform.
Although significant differences were found when examin-
ing the sites as whole, a vegetation type comparison showed
fewer differences. The only difference found between
ICE
and
Pix4D was woody vegetation (Fisher’s exact p = 0.001) (Table
6). Reed canary grass and woody vegetation were found to
be significant between Photoscan Pro and Pix4D (Fisher’s
exact p = 0.011 and 0.007, respectively). This indicates that
the primary difference in performance between the software
packages is in the way they handle imagery acquired over
reed canary grass and woody vegetation.
Unlike image artifacts, which showed no significant dif-
ference with regards to the type of vegetation present on the
ground, image blur was shown to more likely occur with cer-
tain vegetation types. Although no significant difference was
found between the vegetation types in
ICE
(overall comparison
p = 0.065), Pix4D, and Photoscan Pro showed that woody
vegetation preferentially produced more sites with image blur
(Table 7). Again, due to the conservative nature of Bonferroni
corrections the
p
value of 0.017 found between wet-mesic
prairie and woody vegetation in Photoscan Pro was consid-
ered significant. This indicates that tree canopies created
problems for both software programs. One possible explana-
tion for this increased presence of image blur in regions with
woody vegetation is that those regions would receive a greater
amount of windblown movement, which causes leaves to
change both location and orientation between subsequent im-
ages in a flight sequence.
Pricing
The pricing of software licensing ranged greatly from free
to $8,700 USD. Note all prices are in US dollars and were
obtained using public websites in June 2015. Microsoft’s
ICE
is freely available, which is appealing to both business and
educational users
redmond/projects/ice/
). However, Microsoft does not provide
any formal support for this software, although the developers
of this software do participate in the support forum. Agisoft’s
Photoscan Pro is $3,499 USD for a stand-alone license with
education pricing available ($549 USD for a stand-alone
license) and includes 12-months of e-mail support (
http://
/
). Photoscan Pro’s licenses
are perpetual without time-limits. Pix4d Pro Mapper is the
most expensive of the software packages. A perpetual license
is $8,700 USD, 1-year license $3,500 USD, and one-month
license is $350 USD. The perpetual license includes one year
of support with additional support costing $870 USD per year
/
). It should be noted that Pix4d
Mapper Pro license can be installed on two computers to
facilitate in-the-field rapid check as well as lab-based full pro-
cessing. Pix4d also has perpetual education licenses ($1,990
USD) and non-commercial licenses ($4,990 USD). It should
also be noted that reduce pricing of Pix4D is available as part
of some
UAS
packages.
Conclusions
As
sUAS
become an increasingly popular platform for digital
aerial photography, it is important to consider the tradeoffs
between different mosaicking software programs. This re-
search has presented three possible software packages well
suited to
sUAS
digial aerial photography and offer a highly
automated approach: Photoscan Pro, Pix4D, and Microsoft
Image Composite Editor. A comparison was made based on
geometric accuracy, visual quality, and pricing. A ranking
summary of the software in each category can be seen in Table
8.
ICE
produced the fewest visual errors, but had the worst
geometric accuracy. Pix4D and Photoscan Pro were statisti-
cally similar in terms of geometric accuracy, however Pix4D
had better visual quality. It is important to note that all three
software packages struggled with tree canopies where wind-
blown leaf movement led to increasing amounts of image
blur. Overall, this research suggests that, although there is no
single best software option for optimizing all criteria, unless
sub-decimeter geometric accuracy is required for the given
analysis,
ICE
provides the most cost effective option photo-
grammetric applications, although the lack of a DSM product
may deter some users.
T
able
8. O
verall
C
omparison
between
the
S
oftware
P
ackages
. N
umbers
R
epresent
the
S
oftware
O
rder
for
that
C
ategory with
1
being
the
H
ighest
.
T
ie
V
alues
,
indicated
by
“*”,
were
A
warded
in
C
ases where
N
o
S
tatistical
D
ifference was
O
bserved
.
Category
Photoscan Pro
Pix4D
ICE
Geometric accuracy
1.5*
1.5*
3
Visual quality
3
2
1
Cost
2
3
1
Ease of use
1
3
2
References
Agisoft Online Store, 2015. URL:
store/
(last date accessed: 27 April 2016).
Anderson, K., and Gaston, K.J., 2013. Lightweight unmanned aerial
vehicles will revolutionize spatial ecology,
Frontiers in Ecology
and the Environment
, 11(3):138-146.
424
June 2016
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
