Abnormal event detection in indoor video using feature coding

Abnormal event detection in surveillance systems has many applications such as building security, traffic analysis and nursing care. There is a crucial need to investigate the robust and fast methods with high performance for anomaly detection. In this work we used the result of current related methods for anomaly detection regardless of any prior assumption about normal or abnormal events. In this article we have been focused on the unsupervised computer vision algorithm in dynamic scenes. Essentially, the given approach uses a dictionary (basis set) with a completely unsupervised dynamic sparse coding to be adapted to specific data for abnormal events detection. Experimental results on entrance and exit surveillances cameras of subway stations show that the proposed method outperforms other powerfull methods in the literature.