발간년도 : [2022]
| 논문정보 |
|
| 논문명(한글) |
[Vol.17, No.2] A Design of Anomaly Behavior Detection System Based on Deep Learning Model Using CCTV |
|
| 논문투고자 |
Yong Ju Lee |
|
| 논문내용 |
CCTV is undoubtedly the field that produces the most video data and uses the video analysis technology the most. In the mean time, CCTVs have been operated mainly for the purpose of crime prevention or control. In particular, as smart city projects have been promoted in recent years, the number of CCTVs operated by public institutions is increasing. As the number of CCTVs for crime prevention and control increases, the problem of a shortage of control personnel is occurring. In this way, a lot of human resources are required to continuously analyze the video stream collected from CCTV for crime prevention and control purposes, and automation research to solve the labor-intensive camera monitoring problem has become inevitable. To overcome this problem, this paper implements a system that detects abnormal behavior in CCTV images using deep learning. As a method for detecting abnormal behavior, a 2-class classification method that divides abnormal behavior and normal behavior using the UCF-Crime dataset is applied and tested. Behavior patterns are inferred by inputting data that has undergone preprocessing of video data to the trained model. Through the experiment, it is possible to obtain the effect of responding to abnormal behavior patterns by classifying the abnormal behavior patterns based on the deep learning system-based reliable judgment rather than the artificial judgment of the controllers. In the future, it is possible to study specific behavior pattern analysis through diversification of detailed behavior patterns and reliability improvement algorithms of the inference model. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
