基于联合稀疏区域生长的遥感飞机检测算法

Remote Sensing Aircraft Detection Algorithm Based on Joint Sparse Area Growth

针对高分遥感图像中的飞机目标,文章提出了一种基于改进FastMBD和联合稀疏区域生长的精确目标检测算法。首先,基于遥感图像的特性,利用局部区域显著性对FastMBD算法进行改进,克服其先验假设不足;接着,利用内外显著比值实现显著目标检测;最后,基于联合稀疏提出了一种区域生长算法,补全因目标分割导致的残缺,同时基于HOG与SIFT特征,实现了最终目标检测。将提出方法在DOTA数据集上进行实验,结果表明了提出的方法的优越性。

Aiming at aircraft target in high resolution remote sensing image,an accurate target detection algorithm based on improved FastMBD and joint sparse area growth is proposed in this paper.First of all,based on the characteristics of remote sensing image,the FastMBD algorithm is improved by using the local region saliency to overcome the deficiency of prior hypothesis.Then,the saliency target detection is realized by using the internal and external saliency ratio.Finally,an area growth algorithm based on joint sparse is proposed to complete the defects caused by target segmentation.At the same time,the final target detection is realized based on HOG and SIFT features.The proposed method is tested on DOTA data set,and the results show the superiority of the proposed method.