可见光遥感图像中舰船目标检测的多阶阈值分割方法

摘要

针对遥感图像中海洋复杂背景下的海上舰船目标检测,提出了基于自适应多阶阈值分割舰船目标检测方法.使用该方法对Spot,Quickbird,Ikonos,Landsat等不同卫星数据来源、不同分辨率、不同内容的12 704个样本进行了试验,成功检测了1 104个目标中的1 098个,检测率达到99.5%,同时虚警率保持在较低水平.实验证明,该方法对遥感图像中的海上舰船目标检测适应性强、检测率高、稳定性好.%A new multi-level threshold segmentation method was presented for ship detection. The method was designed to detect ship targets in remote sensing images, especially for complex sea surfaces. An experiment on over 12 704 samples, from Spot, Quickbird, Ikonos, Landsat, shows that ship target detection probability can be as high as 99. 5％ while the false alarm rate is low. Experiments testify the method's robustness for various remote sensing data.