| Title |
Abandonment Behavior Detection Using Ganerative Advesarial Networks |
| Authors |
정광희(Kwanghee Jeong) ; 서기성(Kisung Seo) |
| DOI |
https://doi.org/10.5370/KIEE.2021.70.9.1331 |
| Keywords |
Abandonment Behavior Detection; Ganerative Advesarial Networks; Anomaly Detection |
| Abstract |
Abandoned luggage items in public areas will be a potential threat caused by bombs or biological warfare. We present a method for real-time automatic detection of abandoned luggage in video captured by surveillance cameras. Most works for abandoned objects detection use a preliminary step to detect foreground regions or objects, and use various techniques including deep learning to distinguish between abandoned luggage items and other objects. These object detection-based methods require direct learning of characteristics of different types of objects. However, it is difficult to detect a new type of unlearned object, and there is insufficient detection of the instantaneous situation in which an object is abandoned. In order to solve these problems, we propose reconstruction-based anomaly detection to identify unusual patterns that do not conform to expected behavior such as abandoning objects. The approach is comprised of two stages - first, static object detection based on YOLOv4 and second, abandoned luggage recognition based on anomaly detection using Ganerative Advesarial Networks (GAN). We demonstrate the proposed GAN-based anomaly detection for abandonment behavior by performing experiments on Abnormal Event CCTV Video Dataset. |