Moving ship detection based on visual saliency for video satellite
Haichao Li, Yiyun Man · 2016
Moving ship detection is an important issue with the development of video satellite. However, it is difficult for the registration of sea scenes imaging with motion camera. In this paper, we propose a new method to detect moving ship by combining optical flow and video attention saliency for video satellite with image registration. Video visual attention is a consequence of some saliency features such as optical flow, Gabor features, intensity. The model proposed for ship detection mainly includes three stages. Firstly, the moving region is estimated based on optical flow with corners. Secondly, Gabor filter is used for texture features extraction of video images. Finally, the above saliency features as several channels are integrated to a quaternion, which can indicate where the ships are located. The experimental results show that the proposed model can effectively extract moving ships in video images without image registration.