An anomaly-introduced learning method for abnormal event detection

Abnormal event detection aims at identifying anomalies under specific scene and it is widely utilized in health monitoring, public security and pedestrian surveillance. The main challenges are spatiotemporally localizing abnormal and limiting the time cost. Besides, most existing methods only use normal event in training video sequences. We propose an anomaly-introduced learning (AL) method to detect abnormal events. A graph-based multi-instance learning (MIL) model is formed with both normal and abnormal video data. A set of potentially abnormal instances and a coarse classifier are generated by the MIL model. These instances are adopted for an improved dictionary learning, which we call anchor dictionary learning (ADL). The sparse reconstruction cost (SRC) is selected to measure the abnormality. Compared with other methods, we (i) make use of abnormal information and (ii) prune testing instances with a coarse filter and reduce time cost of computing SRC. Experiments demonstrate the effect of our proposed AL method by competitive performance.