Wild animal detection using discriminative feature-oriented dictionary learning

Wild animal detection is an active research area since last many decades among wildlife researchers to study and analyze wild animals and their behavior. This paper presents sparse representation based wild animal detection system using Discriminative Feature-oriented Dictionary Learning (DFDL). DFDL extracts discriminative class-specific features and shows a low complexity method for animal detection. We acquired class-specific dictionaries allowed to represent a new image to identity of the class of the image. Concurrently, these dictionaries are incapable of representing the samples of other classes. The experiments are performed over in-house database compiled by us. We achieved promising results using DFDL with 93% accuracy.