Detecting the degree of anomaly in security videos by using a spatio-temporal feature of change

We propose a method that can discriminate anomalous image sequences for more efficiently utilizing security videos. To match the wide popularity of security cameras, the method is independent of the camera setting environment and the contents of the videos. We use the spatio-temporal feature obtained by extracting the areas of change from the video. To create the input for the discrimination process, we reduce the dimensionality of the data by PCA. Discrimination is based on a 1-class SVM, which is a non-supervised learning method, and its output is the degree of anomaly of the sequence. The method is applied to videos that simulate real environments and the results show the feasibility of determining anomalous sequences from security videos.