Conservational Prospects of Computer-aided Visual Bontebok Identification Using Biometrical FeaturesTilo Burghardt, Anja Wasilewski, Conservational Prospects of Computer-aided Visual Bontebok Identification Using Biometrical Features. 6th International Zoo and Wildlife Research Conference on Behaviour, Physiology and Genetics, pp. 248–250. October 2007. No electronic version available.
Bontebok (Damaliscus pygargus pygargus) are listed as a rare species though their population has recovered from a bottleneck of approximately 70 individuals to an estimated census of 2,300 (in 2005), comprising subpopulations of 4 to 400 individuals. Due to fragmentation and small average herd size, these subpopulations are prone to genetic drift and inbreeding. To preserve the remaining allele variation and enhance genetic fitness, coordinated exchange of breeding stock would appear beneficial; however, both keeping track of individuals during multiple translocations (horizontal gene flow) and collecting pedigree information across several generations (vertical gene flow) pose a challenge.
We argue that biometrical identification can offer a natural, non-intrusive, permanent and expeditiously applicable means for a reliable retrieval of animal identities. Bontebok in particular carry a multitude of individually unique visual features on horns and coat that, used in conjunction, can serve as a biometrical fingerprint. We outline a methodology that can be used to reconstruct bontebok identity on the basis of 2D photographic material combined with a pre-constructed, species-specific computer vision model.
The latter approximates the 3D deformable structure of horns and was built based on 3D depth maps generated by an active stereoscopic laser-camera setup scanning over 12 sets of Damaliscus pygargus horns. It provides both a means for registering the 3D head / horn structure in 2D digital imagery (photos, videos) and a model for estimating the distribution of intra-population variance necessary to infer the performance and robustness of the technique applied.
Such computer aided identification may, in addition to monitoring individuals as horizontal and vertical gene carriers, allow for a wide range of further application. This includes virtual capture-recapture offering a complementary alternative to traditional game counts. It also makes available the option of an individual-based approach for behavioural and ecological studies of free-ranging antelopes on a larger scale.
This can be applied to further the understanding of the social organisation of, for instance, bontebok subpopulations. Exchange of breeding stock involves removal and / or introduction of herd members, affecting the social balance of both the donor and the recipient population and often resulting in increased intra-specific aggression. More detailed knowledge on the stability of social units, social dynamics and interaction patterns derived from individual-based investigations can contribute to minimising socionegative and maximising sociopositive structures within the herds. Preliminary tests with other antelopes indicate that the identification technique should be applicable to a wider range of bovid species.