Professional Documents
Culture Documents
Computer-Aided Photo-Identification System With An Application To Polar Bears Based On Whisker Spot Patterns
Computer-Aided Photo-Identification System With An Application To Polar Bears Based On Whisker Spot Patterns
Key words: capture–recapture, Chamfer, computer vision, image pattern matching, noninvasive identification, reliability,
Ursus maritimus
In studies of population dynamics using capture–recapture threatened or endangered species this method can be
models, estimates of population size and rates of survival, especially undesirable or even prohibited (Kelly 2001;
reproduction, and migration can be obtained only if animals Pennycuick and Rudnai 1970). Additionally, visible tags or
can be identified uniquely (Nichols 1992). Because animals tattoos can interfere with recreational or professional wildlife
often differ in their individual behavior (Fagen and Fagen photography. Marking animals can be convenient, however,
1996; Martin and Kraemer 1987), individual recognition of when handling is necessary for measuring physiological traits,
animals is also essential in studies of behavioral ecology. In sampling blood or tissue, or attaching radiocollars for tracking
analyses of animal movement patterns individual animals are (Ramsay and Stirling 1986).
frequently tracked through space and time (Parks et al. 2006). An alternative method of identifying animals individually is
Thus, many kinds of ecological studies involve identifying through natural markings, such as pelage spot patterns on
individual animals in the field (Nietfeld et al. 1994). cheetahs (Acinonyx jubatus—Kelly 2001), fin marks on
One method of identifying animals individually is by cetaceans (Hammond et al. 1990), and whisker spot patterns
applying artificial markings, such as tags, tattoos, or tissue on lions (Panthera leo—Pennycuick and Rudnai 1970) or
removal (Nietfeld et al. 1994). Artificial markings are unique polar bears (Ursus maritimus—Anderson et al. 2007). The
for each animal, making this method very reliable, unless the main advantage of using natural markings is that it is generally
animal loses its markings (Diefenbach and Alt 1998). noninvasive because the animal can be identified from a
Applying artificial markings, however, is invasive because distance without disturbing it. In addition, using natural
the animal typically must be captured and handled to be markings is practical and affordable because it involves
marked and often recaptured to be identified (Pennycuick and operating accessible equipment (e.g., a photographic camera).
Rudnai 1970). Consequently, this method could be difficult
and expensive to use and possibly detrimental to the health or
behavior of the animal (Kelly 2001; Pennycuick 1978). For www.mammalogy.org
1350
December 2010 ANDERSON ET AL.—AUTOMATED IDENTIFICATION OF POLAR BEARS 1351
The main disadvantage, however, is that unless the entire A reliable identification system should produce similarity
study population is sampled, it is not possible to know for scores for photographs of the same individual that are well
certain whether natural markings of every individual are below the similarity threshold (in this study, we assume that
unique (Pennycuick 1978). In addition, natural markings can the lower the score, the better the match, but this criterion
be difficult to distinguish from a distance, or might be lacking depends on the specific identification system). Two main
altogether, which can lead to incorrect identification (Oliveira- types of error, however, degrade the reliability of an
Santos et al. 2010). Furthermore, searching for an individual’s identification system: photographs of different individuals
identity by visually comparing hundreds or thousands of are incorrectly recognized as a match (i.e., a false positive);
natural patterns can be tedious, error-prone, and time- and photographs of the same individual are not recognized as a
consuming (Hillman et al. 2003; Kelly 2001). match (i.e., a false negative—Hastings et al. 2008; Kelly
To overcome these disadvantages various computer-aided 2001). The 1st type of error, or probability of false positives,
the 2nd image, and uses the mean of these distances as the
final distance between the images. Note that ‘‘nearest’’ does
not imply nearby because the nearest black pixel of the 2nd
image could be relatively far from the 1st and therefore will
have a large Euclidean distance that will increase the overall
Chamfer distance between the images. Also, note that if 2
black pixels are closely aligned, the distance used in the
Chamfer calculation is very small (0 for a perfect alignment),
regardless of the presence of other black pixels. However,
when 2 images come from the same individual but 1 of the
images contains spurious pixels (i.e., black pixels that are not
Poor-angle photographs caused our system to misalign the because the user would have to verify only about 20 individuals
images while computing the similarity score because polar bear on average before finding the correct one. In contrast, if our
whisker spot patterns were foreshortened. In a computer-aided database was large—for example, 1,000 individuals—we might
identification system for whale sharks (R. typus) based on skin tolerate a 5% probability of false positives at the cost of a lower
spot patterns, foreshortening also produced mismatches when probability of true positives. We then can use the relationship
using photographs with angle .30u (Arzoumanian et al. 2005). between the similarity threshold and the probability of false
Nevertheless, using poor-angle photographs is useful because positives to determine the similarity threshold we need to use.
having a record of an individual at various angles improves its We recommend that future evaluations of identification systems
chance of being recognized in the future (Arzoumanian et al. show the relationship between the probability of true positives
2005; Hillman et al. 2003; Kelly 2001). and probability of false positives, or at least report the
Polar bear photographs with poor quality caused smaller probability of false positives alongside the reported probability
whisker spots to disappear during preprocessing, effectively of true positives at the chosen similarity threshold.
changing the extracted pattern. Our method of repeatedly We found that most excellent- and moderate-quality
applying adaptive thresholding to improve robustness was photographs were taken at a distance ,50 m from the polar
most effective with moderate-quality photographs. Objective bears. Anderson et al. (2007) found that whisker spots of polar
definitions of photographic quality are difficult to compare bears were most distinguishable in photographs taken ,50 m
across studies, but others also have found that using poor- away from the bears. Future improvements in digital
quality photographs reduces the probability of true positives photography and more advanced optical equipment should
(Kelly 2001; Whitehead 1990). allow more-distant animals to be recognized. The current
The appropriate similarity threshold for our system can be distance limitation might preclude the use of our system in
chosen based on the trade-off between the probability of true remote locations where achieving proximity to polar bears is
positives and probability of false positives. For example, we difficult (Anderson et al. 2007).
might tolerate a 20% probability of false positives if our In species lacking stripes or spots authors often argue that
reference library was small—for example, 100 individuals— they can use subtle marks, scars, and other attributes to
1356 JOURNAL OF MAMMALOGY Vol. 91, No. 6
TABLE 1.—Number of matching attempts (from a total of 37) that (Anderson et al. 2007). Adult polar bear whisker spot patterns
placed the correct match between 2 polar bear photographs (of change little from year to year, but it is unknown whether
excellent and moderate angle and quality) at the specified position in whisker spot patterns change with the bear’s maturation or
the sorted list of similarity scores. whether they are similar among related polar bears (Anderson et
Position of correct match No. matching attempts al. 2007). In African penguins (Spheniscus demersus), chest
spots on juveniles are not reliable (T. Burghardt, University of
1 27
2 4 Bristol, pers. comm.), and thus only adults could be identified
3 1 using an identification system. Kelly (2001) found no difference
5 1 between similarity scores of related cheetahs (A. jubatus) and
6 2 scores of unrelated cheetahs, implying that cheetah spot patterns
7 1
were not similar among related individuals.
DIEFENBACH, D. R., AND G. L. ALT. 1998. Modeling and evaluation of NIETFELD, M. T., M. W. BARRET, AND N. SILVY. 1994. Wildlife
ear tag loss in black bears. Journal of Wildlife Management marking techniques. Pp. 147–159 in Research and management
62:1292–1300. techniques for wildlife and habitats (T. A. Bookhout, ed.). Wildlife
DYCK, M. G., AND R. K. BAYDACK. 2004. Vigilance behaviour of polar Society, Bethesda, Maryland.
bears (Ursus maritimus) in the context of wildlife-viewing OLIVEIRA-SANTOS, L. G. R., C. A. ZUCCO, P. C. ANTUNES, AND P. G.
activities at Churchill, Manitoba, Canada. Biological Conservation CRAWSHAW, JR. 2010. Is it possible to individually identify
116:343–350. mammals with no natural markings using camera-traps? A
ECKHARDT, G. H. 2005. The effects of ecotourism on polar bear controlled case-study with lowland tapirs. Mammalian Biology—
behavior. M.S. thesis, University of Central Florida, Orlando. Zeitschrift für Säugetierkunde 75:375–378.
ECKHARDT, G. H., J. M. WATERMAN, AND J. D. ROTH. 2002. The PARKS, E. K., A. E. DEROCHER, AND N. J. LUNN. 2006. Seasonal and
functional significance of play fighting in polar bears: are they annual movement patterns of polar bears on the sea ice of Hudson
asocial? Integrative and Comparative Biology 42:1224. Bay. Canadian Journal of Zoology 84:1281–1294.
5. The histogram of the image is modified using histogram Adaptive thresholding (pseudocode)
matching to match the histogram of a good-quality image of Input: integer array A (grayscale image); integer r (neighbor-
polar bear whisker spots. hood radius); integer c
6. Logarithmic transformation (base e) is applied to the image. The Output: an integer array B (black-and-white image)
standard logarithmic function is p 5 c log(1 + p), where c 5 255/ set B 5 integer array with dimensions of A
log(1 + max) and max is the highest pixel value of the image. We for i 5 1, …, width of A
modified the function to account for the minimum value of an for j 5 1, …, height of A
image not always being 0, and we used p 5 c log[(1 + p)/(1 + set mean 5 mean of Ai,j neighborhood (r)
min)], where c 5 255/log[(1 + max)/(1 + min)] and min is the if Ai,j , mean 2 c then
lowest pixel value of the image. set Bi,j 5 black
7. Neighborhood averaging is applied to the image with a else
neighborhood radius of 2. set Bi,j 5 white