Abnormal Activity Recognition Based on Deep Learning in Crowd

Abnormal activity recognition is considered as the most challenging task in surveillance videos. Due to the traditional method depend on the computation of artificial features, and noise data has some influence on the extracted features. In this paper, a new hybrid deep learning structure was proposed to fuse the extracted features, which integrates convolutional neural network (CNN) and long short-term memory network (LSTM). Firstly, the video was preprocessed and extracted visual features by CNN. Next, LSTM was used to learn the temporal features of visual features and added attention mechanism to select important features. Finally, the video feature vector obtained layer by layer to judge abnormal activity. An experiment is used to test the ability of the model on the standard dataset UMN to recognize abnormal activity, the result shows that our experimental demonstrate high performance of recognition and outperform the state-of-art algorithms.