Environ Monit Assess (2019) 191:406
https://doi.org/10.1007/s10661-019-7518-9
EventFinder: a program for screening remotely captured
images
Michael Janzen · Ashley Ritter ·
Philip D. Walker · Darcy R. Visscher
Received: 22 August 2018 / Accepted: 3 May 2019
© Springer Nature Switzerland AG 2019
Abstract Camera traps are becoming ubiquitous
tools for ecologists. While easily deployed, they
require human time to organize, review, and classify
images including sequences of images of the same
individual, and non-target images triggered by envi-
ronmental conditions. For such cases, we developed
an automated computer program, named EventFinder,
to reduce operator time by pre-processing and clas-
sifying images using background subtraction tech-
niques and color histogram comparisons. We tested
the accuracy of the program against images previ-
ously classified by a human operator. The automated
classification, on average, reduced the data requiring
human input by 90.8% with an accuracy of 96.1%,
and produced a false positive rate of only 3.4%. Thus,
EventFinder provides an efficient method for reducing
the time for human operators to review and classify
images making camera trap projects, which compile
a large number of images, less costly to process. Our
testing process used medium to large animals, but
will also work with smaller animals, provided their
images occupy a sufficient area of the frame. While
our discussion focuses on camera trap image reduc-
tion, we also discuss how EventFinder might be used
M. Janzen () · A. Ritter · P. D. Walker · D. R. Visscher
The King’s University, Edmonton, Alberta, Canada
e-mail: Michael.Janzen@kingsu.ca
D. R. Visscher
e-mail: Darcy.Visscher@kingsu.ca
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
in conjunction with other software developments for
managing camera trap data.
Keywords Background subtraction · Camera traps ·
Computer classification · Data reduction ·
Image processing
Introduction
Remote cameras are increasingly important tools for
ecologists from site-specific studies to global monitor-
ing initiatives (Meek et al. 2014; Burton et al. 2015;
Steenweg et al. 2017). While they collect field data
with minimal human attention, the resulting images
require organization, post-processing, and character-
ization before analysis, requiring input of time and
money. Harris et al. (2010) highlighted two general
issues facing camera-aided studies, namely the lack
of systematic organization and problems involving the
volume of imagery produced.
While a number of applications exist to organize,
manage, and facilitate characterization of camera trap
images (Fegraus et al. 2011; Krishnappa and Turner
2014; Bubnicki et al. 2016; Niedballa et al. 2016),
fewer options exist for pre-processing images to
reduce the need for human post-processing. The need
for pre-processing is due, in part, to a high incidence
of non-target images produced by the camera or long
sequences of the same animal. These images create an
“analytical bottleneck” during post-processing (Harris
 406 Page 2 of 10
et al. 2010; Spampinato et al. 2015) and less human
involvement in pre-processing images may help alle-
viate it. In particular, the ability to remove “noisy”
images resulting from background movement and the
ability to identify a single image from a sequence for
classification will greatly reduce processing time.
A primary technique to discriminate a “noisy”
image from an event (animal) is through background
subtraction (Piccardi 2004). This technique helps dif-
ferentiate images containing background noise from
images containing events of interest, and is imple-
mented in video imaging studies (Desell et al. 2013;
Goehner et al. 2015; Swinnen et al. 2014; Weinstein
2015). Camera trap images, as opposed to video, are
separated by longer periods of time making back-
ground subtraction more challenging. We developed
a stand-alone computer program to aid in image pro-
cessing, categorization, and data reduction using back-
ground subtraction to identify candidate foreground
areas and color histogram comparisons to determine
foreground areas from background. The program pre-
processes remotely captured images into relevant and
irrelevant sets for human consideration, identifying
unique events. In this paper, we describe the develop-
ment and procedure of EventFinder, and provide an
assessment of EventFinder based on a known dataset
characterized completely by a human.
Methods
Study area
From June 2015 through June 2016, we maintained
nine cameras (Reconyx PC500) throughout the Cook-
ing Lake—Blackfoot Provincial Recreation Area,
located 45 km east of Edmonton, Alberta. Cameras
were located along trails frequented by animals at
approximately 1.25 m from the ground and set to
collect a burst of three images when triggered. Dur-
ing the time they were active, the individual cameras
collected, on average 3514 (SD = 3214) images each.
Program description
EventFinder pre-processed camera trap images and,
for each sequence of images, determined if the
sequence contained a target animal or noise. For
sequences with a target animal, EventFinder marked
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Environ Monit Assess (2019) 191:406
the image it determined most relevant for human
processing. EventFinder functions as a collection of
smaller programs and modules, each with a focused
task. The program overview is shown in Fig. 1 with
programs and the information passed to the next pro-
gram depicted. Segmenting tasks helps to focus each
programs’ inputs and output, but the process is auto-
matic, so the user needs only to start the program
without concern of which module to start next. In
this respect, the user selects the input images to pro-
cess, optionally specifies parameters, and EventFinder
copies the relevant processed images to a direc-
tory for human classification and identification. We
describe our usage of EventFinder with default set-
tings, but users may set their own parameters and
modify XML instruction files to make customizations
to EventFinder, if they consider a different sequence
of operations provides better image classification.
As labeled in Fig. 1, in our study, the Picture Pro-
cessor program produced sets of images based on
temporal proximity, separating sequences when more
than 60 s had elapsed between sequential images. This
part of the program required the image file names to
link to the images and the time difference between
image sets as input. EventFinder extracts the time the
picture was taken from the file metadata. Users select
files using a GUI (shown in Appendix 1). The output is
a batch file that repeatedly launches Automate Finder,
once for each image set. Automate Finder loads a
default XML file specifying a series of image opera-
tions, which are passed to Find Animal. If Find Animal
is run in user mode, rather than the regular automated
mode, the user can press buttons to initiate image-
processing operations (shown in Appendix 1). User
mode is intended to test different image-processing
sequences to include in the automated mode rather
than use for actual batch image processing since the
user must be present to initiate each operation in this
mode. User mode lets a user graphically test sequences
of operations by clicking buttons rather than requiring
users to enter instructions in an XML file.
Our default image-processing operations for Find
Animal were as follows. To reduce noise and improve
processing speed, Find Animal down-sampled each
image by averaging 16 pixel blocks. The result was
a mapping from 4 × 4 blocks of pixels to one pixel
with less noise. From the images in a sequence, Find
Animal generated a background image and identified
foreground regions. With video frames, a background
Environ Monit Assess (2019) 191:406 Page 3 of 10 406

Fig. 1 The flow, inputs,

and relationship between
components of EventFinder.
Input parameters and
operations are shown in
ellipses while program
modules are shown in
rectangles. The user
specifies inputs into Picture
Processor and the program
calls Automate Finder and
Find Animal to determine
which image for humans to
examine. Providing the
generated CSV file to
Picture Mover moves
retained images to a
separate directory for
human processing

image can be generated with temporal averaging or
other techniques for moving backgrounds (Power and
Schoonees 2002; Van Droogenbroeck and Barnich
2014); however, camera images of an event consisted
of as few as three images causing ghosting, where
foreground objects are included in the background as
faint regions. Thus, when using camera traps instead
of video, only pixels that are actually background
should be included. EventFinder determines which
pixels to include by comparing the pixel in question
with the average value from all pixels at that location,
averaging a location across images in the sequence.
To create a suitable background image, the Find
Animal program computed the standard deviation in
color for each input pixel location across input image
(an overview of background creation is shown in pseu-
docode in Fig. 2). The average standard deviation was
initially used as the threshold to classify foreground
and background pixels to give an initial approxima-
tion of foreground and background pixels. At each
pixel location, two sets of pixels were created, the set
of foreground pixels and the set of background pix-
els. Considering each input image, the pixel at the
location was either included or excluded, forming a
binary label. Connected foreground pixels with the
same binary labels were grouped into regions, sim-
ilar to the approach mentioned in McIvor (2000),
where each region corresponds to a moving object or
noise. The largest region should most frequently cor-
respond to the largest moving animal in the image,
while other regions correspond to additional animals
or noise. EventFinder increased the threshold setting
until the largest region decreased in pixel count by
10%. This value enabled removal of smaller regions

Fig. 2 Pseudocode
highlighting the steps used
to create a background
image

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 406 Page 4 of 10
corresponding to noise, but left the majority of pix-
els for regions corresponding to animals since they
exceeded the threshold by a larger amount than the
noise.
In some situations, as happens when animals over-
lap in multiple images, the majority of pixels at a
location correspond to animal rather than background.
This happens when an animal lingers at a location
and results in a background image with a partial ani-
mal, as shown in Fig. 3. The input images are shown
in Appendix 2. Find Animal corrects these region
misidentifications by checking each potential fore-
ground region against known background locations.
Fig. 3 Two methods of background creation made from an
input image sequence of three input images with temporally
overlapping regions of animal. The first method a uses stan-
dard deviation to identify the background but retains parts of
the animal due to overlap in the majority of the input images.
In the second method, b EventFinder considers nearby known
background areas to select the minority of pixels at locations
representing the actual background, resulting in a truer back-
ground image. The remaining animal fragment seen in the right
of the image was present in all input images. Input images and
masks are shown in Appendix 2
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Environ Monit Assess (2019) 191:406
The logic is as follows, for some pixel locations, all
pixels are within the threshold regardless from which
image in the set they originate. These pixels are con-
sidered definite background and correspond to areas
with no animal at that location in any input image.
Other pixel locations have some pixels within the
threshold and other pixels exceeding the threshold
representing background and foreground respectively.
When an animal lingers in a location, the corre-
sponding pixels containing the animal are incorrectly
considered background. Thus, for each potential fore-
ground region, Find Animal considered the two possi-
ble background regions: one constructed from pixels
with deviations within the threshold and the fragment
constructed from the remaining pixels with deviations
greater than the threshold. Find Animal compared both
region color histograms (histograms of RGB values)
to nearby areas of definitive background and retained
the region with less difference. Consequently, Find
Animal kept the region with fewer pixels as back-
ground when its color closely matched nearby known
background.
For example, in a set of five images containing
a slow moving ungulate, the ungulate may overlap
in three images. This area of overlap constitutes the
majority of pixels coming from three of five images.
Nearby regions contain true background regions con-
sisting of pixels coming from all five images where the
animal did not pass. The color histogram of the pixels
coming from the remaining two images more closely
match true background regions, and consequently, the
background region where the animal slowly crossed is
made from pixels coming from the two images con-
taining the background, rather than the three images
containing the ungulate.
After constructing a proper background, Find Ani-
mal performed background subtraction identifying
foreground fragments and thresholded the result to a
binary image; an example is shown in Appendix 2. In
the binary image, each white area of pixels constitutes
a mask corresponding to pixels containing animal in
the original image. Small foreground fragments, 30
pixels or less, were considered noise and ignored. In
daytime images, fragments consisting mostly of green
were removed since they likely corresponded to mov-
ing vegetation. Images where the majority of the pix-
els were deemed to be the result of background move-
ment were removed since these images typically also
Environ Monit Assess (2019) 191:406
resulted from wind in the image. Image-processing
operations of dilations and corresponding erosions
filled in small holes in fragments and merged nearby
fragments similar to McIvor (2000) (see Gonzalez
and Woods 2007 for image-processing operations).
Some fragments corresponding to a single animal
remained separated, such as when a foreground ele-
ment in the input visually divided the animal, such as
when an animal passed behind a tree. In such cases,
EventFinder ranked fragments and logically merged
nearby fragments if the color histograms matched.
Image sets with no large remaining fragments were
removed as noise. In other sets, the image with the
largest foreground fragment was marked for human
analysis, provided the fragment did not touch an
edge. Thus, for each input image set, at most one
image was retained for human classification. Addi-
tionally, EventFinder computed the average directions
of the center of mass of the largest fragment, indicat-
ing whether the animal is moving left or right, and
the number of large fragments per image, giving an
estimate of number of animals.
Program testing
We tested program efficacy versus human classifica-
tion by comparing program-retained images to those
where humans inspected each image, where a positive
event is an animal in the image. While the charac-
terization of images by humans can result in errors
(Harris et al. 2010), we consider the human char-
acterization as “truth.” We considered the program
to have identified the same event if it retained an
image from the same set as the human. Following
pre-processing, the events identified by the program
were compared to truth using a confusion matrix of
binary outcomes (Landis and Koch 1977). We con-
sider three chief metrics for our program. Firstly, the
removal rate is the proportion of images the program
pre-processed but did not require human analysis,
representing reduced time and cost to researchers.
Secondly, we defined the accuracy of the program to
be the sum of true positive events and true negative
events divided by the total number of images. Thirdly,
the false positive rate indicates the images that the
program assesses as needing human attention but do
not contain a true event. We investigated the sensi-
tivity of our efficacy measures to the elapsed time
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Page 5 of 10 406
between images used to determine unique event sets
based on a sequence of images containing the same
individual.
Results
We found that EventFinder removed on average 90.8%
(SD = 4.5%) of images with an accuracy of 96.1%
(SD = 2.8%) of true events (Table 1). Images retained
by the program as events of interest, but not clas-
sified as such by the human operator (false positive
rate), was 3.4% (SD = 3.0%), showing EventFinder
reduces the human time spent on false events. Indeed,
retained images contained approximately two-thirds
unique and independent events.
The sensitivity analysis on the trade-off between
accuracy and the proportion of images removed sug-
gests that EventFinder is robust with respect to the
user-specified length of time that passes between sub-
sequent images before an image is considered part of
a new set. (Fig. 4). The simultaneous high removal
rate and retention of true event images ensures
EventFinder correctly identifies the majority of events
while greatly reducing post-processing time required
by humans.
Discussion
Numerous recent ecological studies employing cam-
era traps as non-invasive methods for collecting data
emphasize their utility in answering ecological ques-
tions (reviewed in Burton et al. 2015), yet camera trap
usage has issues. Harris et al. (2010) highlighted four
problems with the proliferation of images in camera
trap studies, namely (1) slow image classification, (2)
tedious data entry leading to error, (3) struggling to
keep pace with new data acquisition, and (4) inconsis-
tent filing and naming conventions. With EventFinder
and image pre-processing, we reduce the concerns
surrounding the first three problems. The approxi-
mately 90% reduction in human workload enables
researchers to classify only the images needed, reduc-
ing the monotony of data entry by providing humans
with manageable collections of “target” (event) rich
images, and keep pace with incoming data. This is
particularly true in situations where cameras often
 406 Page 6 of 10 Environ Monit Assess (2019) 191:406

Table 1 Classification performance of EventFinder with respect to the number of events characterized by a human for nine camera
locations

Site Total Human Computer True False False True Removal Accuracy False positive
images events events positive positive negative negative rate rate

1 1213 79 96 70 17 9 1117 0.921 0.979 0.015

2 10118 658 982 560 324 98 9136 0.903 0.958 0.034
3 3595 17 389 17 372 0 3206 0.892 0.897 0.104
4 2867 325 419 295 94 30 2448 0.854 0.957 0.037
5 7170 979 1201 879 222 100 5969 0.832 0.955 0.036
6 1491 65 93 48 28 17 1398 0.938 0.970 0.020
7 1174 25 28 18 3 7 1146 0.976 0.991 0.003
8 298 13 16 13 3 0 282 0.946 0.990 0.011
9 3699 167 320 150 153 17 3379 0.913 0.954 0.043
Average 0.908 0.961 0.034
Standard deviation 0.045 0.028 0.030

Results are for when the time delta between temporally adjacent sequences of images is set to 60 s. Accuracy is calculated as (true
positive + true negative)/total images

100

90
Percent

80

70

15 30 60 120
Time interval (sec)

removed accuracy

Fig. 4 The percentage of images removed (gray) and accuracy (white) of EventFinder across a range of values used as the time delta
beyond which temporally adjacent images are considered a new sequence containing a new event

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Environ Monit Assess (2019) 191:406
capture non-target “noisy” images due to background
movement, or when images form large sequences
of the same individual “loitering.” We suspect these
situations are common in ecological research.
Our work complements existing research in cam-
era trap processing. Other computer programs to
pre-process camera trap images identify individuals
or species, but may require human supervision for
training or segmentation (Yu et al. 2013; Agnieszka
et al. 2017; Yousif et al. 2017). Some camera trap
programs were developed for video, increasing the
potential for a good background image for the iden-
tification of foreground objects of interest; however,
the benefits of using video come at cost in the field
including monitoring and frequent battery or mem-
ory card replacement (Swinnen et al. 2014; Weinstein
2015). Muliple papers consider background subtrac-
tion using video as input, but our system differs in
that it is designed to consider input sequences with
as few as three images (Van Droogenbroeck and Bar-
nich 2014; Hofmann et al. 2012; McIvor 2000; Power
and Schoonees 2002). Consequently, the approaches
listed in processing designed for video input are not
directly suitable for camera trap images, such as the
need to determine a changing background over time
(Van Droogenbroeck and Barnich 2014; Hofmann
et al. 2012). Power mentions that with fewer than
two background modes then it is easier to use simple
subtraction of an averaged background image (Power
and Schoonees 2002). Since this is a near ubiquitous
occurrence in our input, our approach builds on sim-
ple subtraction but addresses the issues of few images,
and that the majority of pixels in a region may come
from a foreground animal.
Other existing methods for classification use a
trained neural network to identify individuals (Yu et al.
2013; Yousif et al. 2017). The system from Yousif
et al. (2017) constructs a background and trained a
deep convolutional neural network on 30,000 image
patches of human, animal, and background to achieve
a classification accuracy as high as 95.6%. Agnieszka
et al. (2016), similar to our work, process camera trap
images into sets based on time, creates a background
image, and segments areas containing motion into
regions; however, their paper does not compare their
system to results from human processing, and as such
it is difficult to assess the accuracy of their system. An
alternative system from Agnieszka et al. (2017) uses a
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Page 7 of 10 406
robust principle component analysis, a cascade object
detector, and a support vector machine to detect snow
leopards based on their spots on their coats with an
accuracy of 93.74%, but requires training. Compared
with previous work, EventFinder’s advantages include
being more general since it is not species specific,
testing its performance against a human’s classifica-
tion with an accuracy of ∼ 96%, and EventFinder
does not require training for pre-processing camera
trap images. Consequently, it is well suited as a gen-
eral tool for use in camera trap studies where custom
tools have not been created or trained. EventFinder is
designed to work with camera trap images as opposed
to video for which other systems are designed (Desell
et al. 2013; Goehner et al. 2015; Weinstein 2015). For
more discussion comparing our approach to some of
these video processing systems, see our previous work
(Janzen et al. 2017).
The trade-off between false positives and true pos-
itive events is a concern whenever dichotomous clas-
sification occurs (Landis and Koch 1977). We have
shown EventFinder to be robust to this trade-off.
Prior work shows roughly equal performance in day
or night conditions (Janzen et al. 2017). Depending
on the study in question, researchers are encouraged
to determine if this trade-off is suitable. We envi-
sion situations, involving rare species, for instance,
where researchers may be more willing to accom-
modate false positives to avoid missing rarely occur-
ring true events. While EventFinder does not com-
pletely remove the human operator, it reduces the time
required to classify image data for further analysis. If
we assume, a trained operator classifies 500 images
per hour (approximately the rate of our human oper-
ator) then the ∼ 32,000 images used in this study
required 63 h of work, while after using EventFinder
only ∼ 7 h are required for ∼ 96% accuracy and
“wasted” time spent on false positives is ∼ 2.5 h.
While a camera set up differs between studies, best
practices can help researchers maximize data collec-
tion and minimize post-collection processing (Meek
et al. 2014). Researchers should be aware of the back-
ground, which may form part of the camera’s detec-
tion zone causing “noisy” images containing moving
vegetation rather than animals. While EventFinder
attempts to identify and remove noise from such
images, it cannot completely compensate for poor
camera placement. The poor performance (high false
 406 Page 8 of 10 Environ Monit Assess (2019) 191:406

positive rate) for one of our sites (site 3; Table 1) unlock the potential of camera traps as an ecological
was primarily due to sub-optimal camera placement tool and to more efficiently answer crucial ecological
where the camera was located in a shallow ditch with questions.
abundant long grass.
Camera settings allowing for bursts of images taken
Acknowledgements We thank Alberta Conservation Associ-
with each trigger provide a useful opportunity to cap-
ation, Alberta Parks, and The King’s University for funding
ture multiple images over which a background image and The King’s Centre for Visualization in Science for com-
is created. They can, however, present the researcher putational support. Special thanks to K. Visser who worked
with long sequences of temporally adjacent images on early versions of this program for an undergraduate project
and to D. Vujnovic for useful discussion on remote camera
that do not represent unique events. This presents a
data. Alberta Parks staff provided logistic support in the Cook-
time cost on humans who inspect each image to deter- ing Lake—Blackfoot Provincial Recreation Area, as well as we
mine if an event occurs when only a single image from thank anonymous reviewers for feedback used to improve this
within the sequence need be classified. EventFinder paper.
uses the burst of images to help delineate background
noise from foreground animals, and by retaining a sin-
gle image the program reduces human classification Appendix: 1
of redundant images.
EventFinder was developed as a stand-alone pro- Software availability:
gram to pre-process images but could be incorpo- http://cs.kingsu.ca/∼mjanzen/CameraTrapSoftware.
rated into a workflow involving existing database, html
naming, and file management programs (Fegraus et al. Example instructions for running the program on a
2011; Krishnappa and Turner 2014; Bubnicki et al. Windows computer are also available from this web-
2016; Niedballa et al. 2016). Currently, the classifi- site. Screen captures for starting the program and user
cation of output depends on additional software to mode control are shown in Figs. 5 and 6 respectively.
tag the image with pertinent metadata and summarize
data into forms for statistical analysis; this software
is camera-specific (e.g., Reconyx MapView Profes-
sional, Reconyx, Holmen, WI, USA) or based on
existing open source options (e.g., TRAPPER Bub-
nicki et al. 2016 or camtrapR, Niedballa et al. 2016).
An additional future development for EventFinder
is the automated identification of animal species found
in the image and the ability to differentiate the number
of individuals contained within an image. Currently,
EventFinder correctly identifies events containing a
single individual 74.4% (SD = 11.4%) of the time
versus multiple individuals (M. Janzen, unpublished
data).
In conclusion, we have shown that EventFinder,
an automated program for pre-processing camera trap
images, aides in reducing the analytic bottleneck
identified by Harris et al. (2010) and Spampinato
et al. (2015) that results from the volumes of images
recorded by camera trap studies. The ∼ 90% reduc-
tion in classification time by humans following pre- Fig. 5 Input for an operator starting the program. The operator
possessing, with high accuracy (∼ 96%) and low selects the images to process and can start the program with
loss of true events (∼ 4%), allows researchers to default settings or modify the parameters

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Environ Monit Assess (2019) 191:406 Page 9 of 10 406

Fig. 6 Control panel for experimenting with user mode if the operator wants to deviate from the default options

Appendix: 2

Input image examples and resulting masks are shown

in Fig. 7.

Fig. 7 Three input images with an overlapping animal (a, b, c). The resulting masks from image (b) without color histogram
comparison (d) and using color histogram comparison (e). The generated backgrounds are shown in Fig. 3

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 406 Page 10 of 10
References
Agnieszka, M., Beery, S., Flores, E., Klemesrud, L., Bayrakci-
smith, R. (2016). Finding areas of motion in camera trap
images. In 2016 IEEE international conference on image
processing (ICIP) (pp. 1334–1338).
Agnieszka, M., Beard, J.S., Bales-Heisterkamp, C., Bayrakci-
smith, R. (2017). Sorting camera trap images. In 2017 IEEE
global conference on signal and information processing
(GlobalSIP) (pp. 249–253): IEEE.
Bubnicki, J.W., Churski, M., Kuijper, D.P. (2016). TRAPPER:
an open source web-based application to manage camera
trapping projects. Methods in Ecology and Evolution, 7(10),
1209–1216.
Burton, A.C., Neilson, E., Moreira, D., Ladle, A., Steenweg,
R., Fisher, J.T., Bayne, E., Boutin, S. (2015). REVIEW:
Wildlife camera trapping: a review and recommendations
for linking surveys to ecological processes. Journal of
Applied Ecology, 52, 675–685.
Desell, T., Bergman, R., Goehner, K., Marsh, R., VanderClute,
R., Ellis-Felege, S. (2013). Wildlife@home: combining
crowd sourcing and volunteer computing to analyze avian
nesting video. In Proceedings - IEEE 9th international
conference on e-Science, e-Science 2013 (pp. 107-115).
Fegraus, E.H., Lin, K., Ahumada, J.A., Baru, C., Chandra, S.,
Youn, C. (2011). Data acquisition and management soft-
ware for camera trap data: a case study from the team
network. Ecological Informatics, 6, 345–353.
Goehner, K., Desell, T., Eckroad, R., Mohsenian, L., Burr, P.,
Caswell, N., Andes, A., Ellis-Felege, S. (2015). A com-
parison of background subtraction algorithms for detecting
avian nesting events in uncontrolled outdoor video. In
2015 IEEE 11th International Conference on e-Science
(e-Science) (pp. 187–195): IEEE.
Gonzalez, R.C., & Woods, R.E. (2007). Digital image pro-
cessing (3rd Edition). New Jersey: Pearson Prentice Hall,
Pearson Education, Inc.
Harris, G., Thompson, R., Childs, J.L., Sanderson, J.G. (2010).
Automatic storage and analysis of camera trap data. The
Bulletin of the Ecological Society of America, 91, 352–360.
Hofmann, M., Tiefenbacher, P., Rigoll, G. (2012). Background
segmentation with feedback: the pixel-based adaptive seg-
menter. In 2012 IEEE computer society conference on com-
puter vision and pattern recognition workshops (pp. 38–43).
Janzen, M., Visser, K., Visscher, D.R., MacLeod, I., Vujnovic,
D., Vujnovic, K. (2017). Semi-automated camera trap
image processing for the detection of ungulate fence cross-
ing events. Environmental Monitoring and Assessment,
189(10), 527.
Krishnappa, Y.S., & Turner, W.C. (2014). Software for min-
imalistic data management in large camera trap studies.
Ecological Informatics, 24, 11–16.
Landis, J.R., & Koch, G.G. (1977). The measurement of
observer agreement for categorical data. Biometrics, 33,
159–174.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Environ Monit Assess (2019) 191:406
McIvor, A.M. (2000). Background subtraction techniques. In
Proceedings of image and vision computing conference
New Zealand.
Meek, P., Ballard, G., Claridge, A., Kays, R., Moseby, K.,
O’Brien, T., O’Connell, A., Sanderson, J., Swann, D.,
Tobler, M., et al. (2014). Recommended guiding principles
for reporting on camera trapping research. Biodiversity and
Conservation, 23, 2321–2343.
Niedballa, J., Sollmann, R., Courtiol, A., Wilting, A. (2016).
camtrapR: an R package for efficient camera trap data
management. Methods in Ecology and Evolution, 7, 1457–
1462.
Piccardi, M. (2004). Background subtraction techniques: a
review. In 2004 IEEE international conference on systems,
man and cybernetics, (Vol. 4 pp. 3099–3104): IEEE.
Power, P.W., & Schoonees, J.A. (2002). Understanding back-
ground mixture models for foreground segmentation. In
Proceedings of image and vision computing conference
New Zealand (pp. 267–271).
Spampinato, C., Farinella, G.M., Boom, B., Mezaris, V., Betke,
M., Fisher, R.B. (2015). Special issue on animal and insect
behaviour understanding in image sequences. EURASIP
Journal on Image and Video Processing, 2015, 1.
Steenweg, R., Hebblewhite, M., Kays, R., Ahumada, J., Fisher,
J.T., Burton, C., Townsend, S.E., Carbone, C., Rowcliffe,
M.J., Whittington, J., Brodie, J., Royle, J.A., Switalski,
A., Clevenger, A.P., Heim, N., Rich, L.N. (2017). Scaling-
up camera traps: monitoring the planet’s biodiversity with
networks of remote sensors. Frontiers in Ecology and the
Environment, 15(1), 26–34.
Swinnen, K.R., Reijniers, J., Breno, M., Leirs, H. (2014). A
novel method to reduce time investment when processing
videos from camera trap studies. PloS One, 9, e98,881.
Van Droogenbroeck, M., & Barnich, O. (2014). viBe: a disrup-
tive method for background subtraction. In T. Bouwmans,
F. Porikli, B. Höferlin, A. Vacavant (Eds.) Background
modeling and foreground detection for video surveillance
(pp. 7.1–7.23): Chapman and Hall/CRC.
Weinstein, B.G. (2015). Motionmeerkat: integrating motion
video detection and ecological monitoring. Methods in
Ecology and Evolution, 6, 357–362.
Yousif, H., Yuan, J., Kays, R., He, Z. (2017). Fast human-animal
detection from highly cluttered camera-trap images using
joint background modeling and deep learning classifica-
tion. In 2017 IEEE international symposium on circuits and
systems (ISCAS) (pp. 1–4).
Yu, X., Wang, J., Kays, R., Jansen, P.A., Wang, T., Huang, T.
(2013). Automated identification of animal species in cam-
era trap images. EURASIP Journal on Image and Video
Processing, 2013, 52–61.
Publisher’s note Springer Nature remains neutral with regard
to jurisdictional claims in published maps and institutional
affiliations.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”),
for small-scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are
maintained. By accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use
(“Terms”). For these purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or
a personal subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or
a personal subscription (to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the
Creative Commons license used will apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data
internally within ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking,
analysis and reporting. We will not otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of
companies unless we have your permission as detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that
Users may not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to
circumvent access control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil
liability, or is otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by
Springer Nature in writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer
Nature journal content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates
revenue, royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain.
Springer Nature journal content cannot be used for inter-library loans and librarians may not upload Springer Nature journal
content on a large scale into their, or any other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any
information or content on this website and may remove it or features or functionality at our sole discretion, at any time with or
without notice. Springer Nature may revoke this licence to you at any time and remove access to any copies of the Springer Nature
journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express
or implied with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or
warranties imposed by law, including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be
licensed from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other
manner not expressly permitted by these Terms, please contact Springer Nature at

onlineservice@springernature.com