Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Identification of Great Apes using Gabor features and Locality Preserving Projections

: Loos, Alexander


Association for Computing Machinery -ACM-:
1st ACM International Workshop on Multimedia Analysis for Ecological Data, MAED 2012. Prodeedings : 2 November, Nara, Japan In Conjuction with 20th ACM Multimedia 2012
New York: ACM, 2012
ISBN: 978-1-4503-1588-3
International Workshop on Multimedia Analysis for Ecological Data (MAED) <1, 2012, Nara>
International Conference on Multimedia (MM) <20, 2012, Nara>
Conference Paper
Fraunhofer IDMT ()
animal biometrics; face recognition; AVS

In the ongoing biodiversity crisis many species, particularly primates like chimpanzees for instance are threatened and need to be protected. Often, autonomous monitoring techniques using remote camera devices are used to estimate the remaining population sizes. Unfortunately, the manual analysis of the resulting video material is very tedious and time consuming. To reduce the burden of time consuming routine work, researches have recently started to use computer vision algorithms to identify individuals. In this paper we present an approach for automatic face identification for primates, especially chimpanzees. We successfully combine Gabor features with Locality Preserving Projections (LPP). As classifier we use a new method called Sparse Representation Classification (SRC). In two experiments we show that our approach outperforms a recently published algorithm for face recognition of Great Apes. We also compare our algorithm to other state-of-the-art face recognition algorithms using three methods for feature-space transformation and two different classification approaches, namely SRC and an enhanced version called Robust Sparse Coding (RSC). Our approach not only outperforms the other algorithms for full-frontal faces but is also more invariant to pose changes. For our experiments we use two publicly available, real-world databases of captive and free-living chimpanzees from the zoo of Leipzig, Germany and the Taï National Park, Africa, respectively. Even though both datasets are very challenging due to difficult lighting conditions, non-cooperative subjects, various pose changes and even partial occlusion, the achieved recognition rates are very promising and therefore our approach has the potential to open up new ways in effective biodiversity conservation management.