earticle

영어

In this paper, we propose a noble method of detecting abnormal events for intelligent CCTV service. The method detects the object asynchronously by using object classification and bounding-box estimation, which are based on a light-weight CNN model. To analyze the abnormal events, we categorized four human-related events (loitering, intrusion, abandoned object, and violence) that are classified by time series data analysis. Experimental results showed that the average accuracy of the event detection is 95.8%. We expect that the proposed method can be applied to a real-time commercial intelligent CCTV service.