一种利用SAR和可见光图像融合检测目标的方法

本文利用军事目标在SAR图像中具有较大的雷达散射截面,后向散射强以及在可见光图像中几何外形清晰的特点,提出一种利用SAR图像和可见光图像多维特征检测目标的方法。该方法分为图像预处理,目标检测和融合检测三部分。首先,利用基于特征匹配的方法对多传感器图像进行配准。其次,利用全局双参数恒虚警(CFAR)方法检测SAR图像中的目标,经过滤波处理后,确定感兴趣区域(Region Of Interesting,ROI)并提取目标的SAR图像特征；将ROI映射到可见光图像中,对该区域进行边缘检测、滤波、连通性分析、提取目标的可见光图像特征。最后,在特征层利用特征向量距离准则融合检测目标。实验结果表明该方法性能优于单传感器检测方法,且能有效的改进目标检测性能。 

A Fusion of Target Detection Method from SAR Image and Optical Image

A new targets detection method is proposed in this paper, which take advantage of the composite properties of targets in synthetic aperture radar(SAR) image and optical image, based on that military targets have strong radar cross section(RCS), strong back scattering characteristics and easy to detect in SAR images and the targets in the optical images have good geometry features and easy to distinguish the targets from background. There are three steps about the approved method: images preprocessing, target detection and data fusion on features. At first, the images from multisensor are registered with feature based registration algorithm. The second, the global Constant False Alarm Rate(CFAR) algorithm is used to detect by targets in SAR images, after filtering processing, the Regions of Interest(ROI) are obtained in SAR image. Mapping the ROIs to optical image, the edge features and characteristics of connected areas of targets are extracted in the ROIs of optical images. Finally, the distance of feature vectors is used to detect targets on feature level. The experimental result shows this algorithm has better performance than detection algorithm on single sensor.