We propose an anomaly detection approach by learning a generative model using deep neural network. A weighted convolutional\nautoencoder- (AE-) long short-termmemory (LSTM) network is proposed to reconstruct rawdata and performanomaly detection\nbased on reconstruction errors to resolve the existing challenges of anomaly detection in complicated definitions and background\ninfluence. Convolutional AEs and LSTMs are used to encode spatial and temporal variations of input frames, respectively. A\nweighted Euclidean loss is proposed to enable the network to concentrate on moving foregrounds, thus restraining background\ninfluence. Moving foregrounds are segmented from the input frames using robust principal component analysis decomposition.\nComparisons with state-of-the-art approaches indicate the superiority of our approach in anomaly detection. Generalization of\nanomaly detection is improved by enforcing the network to focus on moving foregrounds
Loading....