A Unified Framework for Joint Moving Object Detection and Tracking in the Sky and Underwater

Abstract

AbstractThe ability to detect and locate the moving object in a video is a fundamental procedure in applications of computer vision. However, these tracking methods still face some challenges, and are contradictory among different tasks. In this paper, a unified framework for joint moving object detection and tracking in the sky and underwater is proposed. This framework meets the requirements of two real applications: (i) tracking unmanned aerial vehicle (UAV) in the sky; and (ii) tracking unmanned underwater vehicle (UUV) in water. It consists of three key steps: (i) moving object detection by pixel classification; (ii) data association by blob detection; and (iii) object tracking by efficient convolution operator. Finally, analysis on the accuracy of the proposed framework is provided. Experimental results on real-world datasets and object tracking benchmark (OTB) demonstrate the advantage of the tracking method compared with some state-of-the-art trackers, in terms of accuracy and robustness. In addition, to the best of the authors’ knowledge, there is no previously published work for joint moving target detection and tracking in the sky and underwater.KeywordsBackground subtractionTrackingPixel classificationUnmanned aerial vehiclesUnmanned underwater vehicles