目前的目标融合检测方法大都是基于多源遥感图像配准的,然而在实际的应用中,成像机理不同的多源遥感图像的精校正和图像间的配准是十分复杂的,难以确保其配准精度.为此,本文提出了一种基于目标关联的多源卫星遥感图像的兵营融合检测方法.该方法不对图像进行配准,而是根据单源图像的目标自动检测结果,利用图像的大地坐标信息,截取包含目标的同一地区的局部遥感图像,再分别提取多源遥感图像目标的特征,并根据其中冗余的特征,对提取的目标区域建立关联,再由关联检验确保特征关联的正确性,最后对目标特征进行融合决策,得到目标融合检测结果.实验结果表明,该方法能有效地利用多源遥感图像的信息,降低遥感图像目标检测的误判率,提高目标特征的准确度.

An effective feature fusion strategy applied in barracks detection is proposed,which takes advantage of the complementary target features in Synthetic Aperture Radar(SAR) and optical satellite images.With more details,it is easier to detect barracks in optical satellite images.However,extracted features in optical satellite images may be unreliable because of the complex background.Thus feature fusion between SAR and optical satellite images is needed. Traditional fusion detection methods are commonly using image registration to set up mapping of the same target in multi-source satellite images.However,due to side-looking character of SAR imaging,the hypsography may cause distortion and precise rectification of the distortion will be very complex.What is more, reference control points are different to select because of the different imaging methods,and co-registration between SAR and optical satellite images may be not perfect.Thus,in this paper we tried to avoid precise rectification and co-registration.The proposed method firstly locates barracks in optical satellite images and cuts the SAR sub images of the same location by adopting the geodetic coordinates information of images.It then independently extracts the feature information of detection results in both optical satellite images and SAR sub images and makes the extracted areas associated according to the redundant feature information.Also a verification method is present to make sure the association is correct.Finally,all the features are fused to get a more accurate and integral description of targets.Results obtained prove that this method can fully use the information of multi-source satellite images,reduce detection mistakes and improve the accuracy of targets features.