Moving Object Detection from a Moving Stereo Camera via Depth Information and Visual Odometry

Abstract

It is important for driving safety to know whether surrounding objects are moving or static. Most existing methods use pre-trained object detectors to detect vehicles and humans before determine whether they are moving or not. However, there are not only these two type objects on the real road. This study presents a system using depth information and visual odometry to detect moving object. It also adopts an adaptive thresholding method to enhance performance on detection. The proposed system can accurately detect moving objects according to the experimental results.