摘要
为了实现岛岸复杂环境下航空侦察图像舰船目标检测,提出一种基于区域协方差和目标度的显著目标检测方法。在联合条件随机场和字典学习的图像显著性检测框架下,首先提取每个区域增强的sigma特征,并进行稀疏编码,然后又设计基于显著优化的目标度特征,利用信念传播算法推断生成舰船目标显著图,最后应用高效子窗口搜索方法实现舰船目标检测。实验结果表明,新方法的显著图结果目标区域一致高亮,背景杂波抑制效果好,可实现准确的目标检测。
In order to realize the ship target detection of the aerial reconnaissance image under complex environment with island and shore, a salient target detection method is proposed based on regional covariance and objectness. Under the saliency detection framework of conditional random field and dictionary learning, the sigma features of each region are extracted and sparsely coded, and then the objectness feature is designed by using saliency optimization and the belief propagation algorithm is adopted to infer the saliency map of the ship target image, and finally the efficient subwindow search method is applied to achieve ship target detection. The experimental results show that the saliency map of the proposed method has complete target features and good background suppression, and it can achieve accurate target detection.
作者
刘松涛
姜康辉
刘振兴
LIU Songtao;JIANG Kanghui;LIU Zhenxing(Department of Information System, Dalian Naval Academy, Dalian 116018, China)
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2019年第5期972-980,共9页
Systems Engineering and Electronics
基金
国家自然科学基金(61303192)
博士后基金(2015M572694
2016T90979)资助课题
关键词
显著性检测
条件随机场
区域协方差
目标度
高效子窗口搜索
saliency detection
conditional random field
region covariance
objectness
efficient subwindow search
