<![CDATA[In recent years, single object tracking has been extensively studied and achieved much development. However, multiple objects tracking is still an issue that remains to be addressed. Generally speaking, existing multiple objects tracking methods employ a manner of simultaneously tracking each object respectively. In this paper, we develop a multiple ship tracking algorithm based on deformable part model to accomplish multiple ship tracking in inland waterway CCTV (Closed-Circuit Television) automated surveillance. Our method utilizes HOG features to construct the appearance models of ships. Then by taking full advantages of the spatial constrains between ships, we can successfully explore mutual relations for multiple ships, thus accomplishing multiple ship tracking in its true sense. Moreover, structured learning method is used to learn how to update the model parameters. Numerous experimental results on challenging inland waterway CCTV video sequences demonstrate that our method can effectively and accurately perform robust multiple ship tracking.]]>
Copyrights © 2014
