基于目标特征的光学遥感图像舰船检测方法

针对在大场景、高分辨率的光学遥感图像舰船目标检测中，传统的阈值检测方法效果不佳，而恒虚警法计算量大的缺点，研究一种根据检测舰船目标的特征，对每个可能的目标区域，快速计算检测阈值的方法。同时，研究基于RNN网络及信噪比的特征选择方法，对目标的特征进行优选并对候选目标进行鉴别。实验结果表明，采用本文方法进行舰船目标检测能获得较好的检测结果.

A Method of Ships Detection From Optical Remote Sensing Images Based on Feature Detection

The traditional methods such as detection by a threshold or CFAR do not very well in ships detection form large scene and high resolution remote sensing images. To overcome these drawbacks, a method based on ship feature detection is proposed in this paper, which can compute threshold of every potential target with high speed. Besides, a method of feature selection based on RNN algorithm and signal noise ratio is proposed, which can not only select good features of a target, but also identify ships from candidate detected targets. A number of experiments demonstrates good performance of the proposed method.