Infrared ship detection is challenging in dense sun-glint scene because some sun-glint clutters are similar to ships in grayscale and local contrast. The large amount of sun-glint clutters increases the false alarm probability of infrared ship detection. Considering that, in infrared sequence images, the ship is stable, while the clutter is constantly changing with the motion of wave. A feature about fluctuation can be designed to describe the changes of candidate targets and eliminate clutters. Meanwhile, in a single-frame infrared image, the ship has a specific shape, while the clutter does not. A high-dimensional feature about structure can be designed to describe the shape of candidate targets and distinguish ships and clutters. In this paper, a novel method based on time fluctuation feature and space structure feature for infrared ship detection in dense sun-glint scene is proposed. The proposed time fluctuation feature and space structure feature can accurately describe the difference between ships and clutters, thereby achieving high-precision and high-recall infrared ship detection in sun-glint scene. Extensive experiments show that the proposed method outperforms the existing methods, especially in dense sun-glint scene.

在浓密的日照场景中，红外船的检测具有挑战性，因为某些日照杂波在灰度和局部对比度方面类似于船舶。大量的日光闪烁杂物增加了红外舰船检测的误报概率。考虑到在红外序列图像中船是稳定的，而杂波却随着波的运动而不断变化。可以设计有关波动的功能来描述候选目标的变化并消除混乱。同时，在单帧红外图像中，船具有特定的形状，而杂波则没有。可以设计有关结构的高维特征，以描述候选目标的形状并区分飞船和杂波。在本文中，提出了一种基于时间波动特征和空间结构特征的密集日照场景中红外舰船检测新方法。提出的时间涨落特征和空间结构特征可以准确地描述舰船与杂波之间的差异，从而实现了日照场景中高精度，高召回率的红外舰船探测。大量实验表明，所提出的方法优于现有方法，特别是在阳光密集的情况下。