General-purpose Abandoned Object Detection Method without Background Modeling

This paper proposes an effective method for detecting abandoned objects in surveillance videos. We use a pedestrian detector trained by YOLO deep learning model to detect surveillance videos. In this process, key frames before and after pedestrians pass through the scene can be obtained. Subsequently, we compare and analyze the key frames to get the position of abandoned objects. It is a general purpose abandoned object detection method that does not use background modeling. The experimental results obtained based on the ABODA database show that this method is effective for detecting abandoned objects and is more robust to illumination changes