Login

Join for Free!
114103 members
table of contents table of contents

This paper assessed the reliability of this simple, freely available software package …


Biology Articles » Conservation Biology » Spot the match – wildlife photo-identification using information theory » Results

Results
- Spot the match – wildlife photo-identification using information theory

I3S (Interactive Individual Identification Software) matching validation
The Information Criterion weights (w) for the most parsimonious matches (w1) for the 50 matched pairs (100 images) were broadly distributed between 0.05 and 0.85, while w1 for the 50 non-matched pairs (100 images) were highly right-skewed (Fig. 3a,b). All w1 for non-matched pairs were w1 for matched pairs was 0.32 (± 0.05), which was much greater than the median for non-matched pairs (0.06 ± 0.01). Evidence ratios for the best-matched relative to the next-highest matched images (ER1) for known matched pairs were also highly right-skewed and ranged from 0.73 to 51.92, with a median of 7.36 (± 2.45) (Fig. 3c). ER1 for non-matched pairs were all 3d). Evidence ratios for the second best-matched relative to the next-highest matched images (ER2) for known matched pairs ranged from 0.73 to 114.18, with a median of 7.57 (± 3.82). ER2 for non-matched pairs were also all

Overall, 93 images out of the 50 known-matched pairs were matched correctly using I3S. w1 for the correctly assigned matches ranged from 0.05 to 0.85 (median = 0.36 ± 0.05), and their ER1 ranged from 0.73 to 51.92 (median = 8.82 ± 2.56) (Fig. 4a,b). Known-matched photographs that I3S failed to match (7 images) had w1 that ranged from 0.05 to 0.14 (median = 0.07 ± 0.02), with their ER1 ranging from 0.95 to 2.28 (median = 1.23 ± 0.36).

Assessing 'by-eye' matches using I3S

Of the 33 individuals re-sighted between years in the database used by Meekan et al. [19], 10 individuals could not be matched with I3S because their images were not amenable to I3S fingerprinting (absence of reference points) or their match was not present in the database. This was because the Meekan et al. [19] study also used images from a separate database and included scar-identified individuals that were not available for photographic matching using I3S. Thus, we could only re-assess 23 of these by-eye matches that included 13 LS matches and 16 RS matches (58 images total).

Forty-eight of the 58 images (83%) from the 23 individuals were matched correctly using I3S. w1 for the correctly assigned by-eye matches ranged from 0.05 to 0.53 (median = 0.16 ± 0.04) (Fig. 5a), and their ER1 were between 1.04 and 24.57 (median = 2.33 ± 1.58) (Fig. 5b). Incorrectly assigned by-eye matches had w1 ranging from 0.04 to 0.13 (median = 0.06 ± 0.01) and their ER1 ranged from 0.67 to 2.76 (median = 1.04 ± 0.37). I3S also identified two images that were false positives (i.e., sharks that were incorrectly matched with other photographs) in the by-eye matching process. Neither of these images was matched with other known images of the identified sharks.

Horizontal angle

Mean w1 decreased linearly as the horizontal angle of subjects within images increased (Fig. 6a). Median w1 ranged between 0.92 (± 0.06) for angles of 10°, to 0.29 (± 0.13) for angles of 40°. The images of subjects at 30° had w1 approaching those of non-matching pairs, and the distribution of w1 for images of subjects at 40° overlapped the distribution of w1 for non-matching pairs (Fig. 6a).

There was an exponential decline of median ER1 with increasing angle (Fig. 6b). Median ER1 ranged from 69.16 (± 52.24) for images of subjects at 10°, to 1.56 (± 2.81) for images of subjects at 40°. The distribution of ER1 for images of subjects at 30° approached that for non-matching pairs, and the distribution of ER1 for images of subjects at 40° overlapped the ER1 distribution for non-matching pairs.

Number of spot pairs

There was evidence for a negative relationship between the transformed I3S scores and spot pairs (ER = 9.94 × 105, adjusted R2 = 0.26; Fig. 7a), but no evidence for a relationship between w1 and the number of spot pairs (ER 7b).


rating: 0.00 from 0 votes | updated on: 25 Apr 2007 | views: 4073 |

Rate article:







excellent!bad…