CAJ | 학술논문

针对海面光学卫星遥感图像舰船检测中云层干扰所产生高虚警率的情况，提出一种基于串并联混合多维特征融合判决的舰船目标检测算法。首先，利用高斯灰度阈值分割目标与海面云层背景，并对分割后的二值图像进行区域标记；然后，利用感兴趣区域的几何形态和边缘能量两类特征，包括中心偏移度、长宽比、面积变化比、曲折度和填充度，构建了串并联混合多维特征融合判决方法；最后，通过对感兴趣区域进行综合鉴别，有效剔除了云层干扰导致的虚警，解决了含云层背景检测结果虚警率高的问题。对多幅SPOT4光学遥感卫星图像的实验结果表明算法的有效性。
침대해면광학위성요감도상함선검측중운층간우소산생고허경솔적정황，제출일충기우천병련혼합다유특정융합판결적함선목표검측산법。수선，이용고사회도역치분할목표여해면운층배경，병대분할후적이치도상진행구역표기；연후，이용감흥취구역적궤하형태화변연능량량류특정，포괄중심편이도、장관비、면적변화비、곡절도화전충도，구건료천병련혼합다유특정융합판결방법；최후，통과대감흥취구역진행종합감별，유효척제료운층간우도치적허경，해결료함운층배경검측결과허경솔고적문제。대다폭SPOT4광학요감위성도상적실험결과표명산법적유효성。
According to problem of high false alarm rate in ship detection produced by clouds in optical remote sensing image,a detection algorithm based on serial-parallel and multi-feature fusion is proposed. First,the segmentation of ships and clouds is realized by Gaussian-gray-threshold,and the regions of binary image are marked.Then,the serial-parallel and multi-feature fusion algorithm is constructed by the geometric morphological and edge-energy characteristics,including center deviation degrees,length-width ratio,area changing ratio,sinuosity and degree of filling.False alarms caused by clouds are effectively eliminated though the comprehensive judgment of region of interesting ( ROI),thus the problem of high false alarming rate is solved.The effectiveness of the algorithm is verified by the experiment results with multiple optical remote sensing satellite images of SPOT4.