基于特征融合的可见光图像舰船检测新方法

A New Method for Ship Detection Based on Feature Fusion in Optical Image

该文以可见光图像舰船目标为研究对象,提出了用多谱图像和全色图像进行特征融合来检测舰船目标的方法。该方法首先利用多谱图像实现水域和陆地的分离,然后把分类结果映射到全色图像上从而实现在全色图像上区分水域和陆地;屏蔽陆地后用Otsu方法分别在多谱图像和全色图像上对目标进行分割,并提取目标特征,最后对目标特征进行融合来检测舰船目标。实验证明该方法有效可行。

Aiming at ship detection in optical image,this paper proposes a new method for ship detection based on feature fusion of multi-spectral image and PAN image.First,it does segmentation work in multi-spectral image and separates sea and land.Next,the previous result is mapped to the PAN image and sea-land separation in PAN image is quickly done.In the following,the Otsu method is adopted to extract objects in two kinds of images.At last,the two groups of features in two different kinds of images are fused to detect ships.The experiment result shows that the new method can detect ships more efficiently and effectively.