Moving Object Detection based on Deep Atrous Spatial Features for Moving Camera

Ajmal Shahbaz, Kang-Hyun Jo

Abstract: Moving object detection in a moving camera is a challenging task. The task is not only challenging due to a change in the pose of a camera. But the challenges posed by moving objects add on such as illumination changes, dynamic background, shadows, camouflaged, and static moving objects. This paper proposed moving object detection using a moving camera. The algorithm is developed using the deep learning technique. The deep dilated features are extracted using a dilated convolution kernel. The features are expanded using a bilinear interpolation based network. The proposed algorithm is compared with the top-ranked algorithms in the moving object detection field on the change detection dataset.