Abnormal human activity detection by convolutional recurrent neural network using fuzzy logic

In automated video surveillance applications, detecting abnormal human activity is incredibly difficult to classify them. The automatic detection of aberrant human activity in a surveillance system was resolved in our proposed work. The videos are first turned into frames, and keyframes from a batch of frames are extracted using fuzzy logic. Secondly, the features are retrieved from the keyframes using a pre-train convolutional neural network (CNN) through transfer learning. Finally, to recognize anomalous activity from video, the collected features are loaded into a Long-Short Term Memory (LSTM) based recurrent network. Two benchmark datasets were used to evaluate the proposed methodology: the UCF50 and the UCF-crime, with our model achieving 95.04% and 49.04% accuracy, respectively. Using the same data set, the experimental findings are compared to conventional detection approaches which suggest that our proposed model outperforms the other approaches that were compared.