Özet:

crowded scenes. A new concept of atomic event is introduced into this framework, which is the basic component of video events. Different from previous bag-of-words (BoW) modeling-based methods that represent feature descriptors using only one code word, a feature descriptor is represented using a few more atomic events in bag-of-atomic-events (BoAE) modeling. Consequently, the approximation error is reduced by using the obtained BoAE representation. In the context of abnormal event detection, BoAE representation is more suitable to describe abnormal events than BoW representation, because the abnormal event may not correspond to any code word in BoW modeling. Fast latent Dirichlet allocation is adopted to learn a model of normal events, as well as classify the testing event with low likelihood under the learned model. Our proposed framework is robust, computationally efficient, and highly accurate. We validate these advantages by conducting extensive experiments on several challenging datasets. Qualitative and quantitative results show the promising performance compared with other state-of-the-art methods.

Anahtar Kelimeler: