基于显著性的遥感图像舰船目标检测

Ship Target Detection Based on Saliency in Remote Sensing Image

舰船作为海洋信息感知中的重要目标,其检测在军舰探测、精确制导等军用领域以及海面搜救、渔船监测等民用领域具有极其重要的战略意义。海洋遥感图像受云雾、风浪、海杂波和光照等干扰使得舰船检测具有挑战性。根据可见光遥感图像舰船目标检测特点提出粗检测和细鉴别相结合的技术路线。先基于视觉显著性的谱残差法对图像进行增强以提取目标候选区域,后根据舰船与干扰因素差异采用舰船方向梯度直方图特征对目标候选区域进行鉴别,提取真正的舰船目标。实验结果表明,上述算法舰船检测率高,对光照、海杂波干扰具有一定程度的鲁棒性,且能有效剔除碎云岛屿等干扰物,显著降低虚警率。

As an important target of marine information perception,ship detection has a very important strategic significance in the military fields such as warship detection,precise guidance,as well as the civilian fields such as sea search and rescue,fishing vessel monitoring and so on.Ocean remote sensing image is challenged by the interference of cloud,wave,sea clutter and light.In this paper,based on the characteristics of ship target detection in optical remote sensing images,the technology of rough detection and fine identification is proposed.Firstly,based on the spectral residual method of visual saliency,the image was enhanced to extract the target candidate area.Then,according to the difference between the ship and the interference factors,the target candidate area was identified by using the ship histogram of oriented gradient(S-HOG)feature to extract the real ship target.The experimental results show that the algorithm has a high detection rate of ships,a certain degree of robustness to the interference of light and sea clutter,and can effectively eliminate the interference such as broken cloud islands,and significantly reduce the false alarm rate.