静止轨道遥感卫星海面运动舰船快速检测方法

摘要

面向静止轨道光学遥感卫星，该文提出一种海上运动舰船目标快速检测方法。该方法首先利用多结构多尺度形态学滤波对海洋背景遥感图像进行背景抑制；然后采用自适应阈值分割和自组织聚类获得候选目标；再根据目标运动特征，利用静止轨道卫星凝视序列图像对候选目标进行多目标移动式邻域判决，剔除虚假目标；最后关联舰船目标以及融合卫星平台数据，可快速计算舰船状态等深层次动态信息。实验结果与分析表明，所提方法能有效检测海洋背景遥感图像中的多个运动舰船目标，准确获取各个舰船位置、航速、航向、运动轨迹等信息，具有算法简单，目标检测率高、虚警率低，稳定性好等优点。该方法为我国静止轨道光学遥感卫星在轨数据处理与应用提供了技术支撑。%A moving ship detection method is presented for ocean moving objects detection of remote sensing satellite in geostationary orbit. First, the multi-structural and multiscale element morphological filter is used to suppress background information of oceanic remote sensing images. Then, image segmentation is done by adopting the adaptive threshold algorithm. Accordingly, the connected domains of pre-detection targets are obtained by utilizing self-organized clustering. Finally, real targets from many candidate targets can be obtained by multi-object variable region decision based on moving targets feature. The experiment results and analysis show that the proposed method can detect moving warship targets and the trajectories of moving targets efficiently, and has high detection probability and robustness. This method provides technical support for on-board image processing of remote sensing satellite in geostationary orbit.