作者: Shifeng Li , Chunxiao Liu , Yuqiang Yang
DOI: 10.1016/J.PATREC.2017.09.001
关键词:
摘要: Abstract In this paper, we propose a novel method to detect abnormal events from videos based on maximum posteriori (MAP). Conventional methods consider the with low-probability respect model of normal behavior as anomaly. Different traditional approaches, anomaly detection is achieved by MAP estimation in our framework. The prior knowledge obtained background subtraction due fact that anomalies often occur at locations consisting moving objects, and likelihood function computed comparing similarity between testing samples designed grid template. Experiments three public databases show can effectively complex scenes.