摘要
船舶检测是SAR海洋应用的重要方面。提出一种通用的检测方法,用以检测不同状况下的SAR图像船舶目标。首先将SAR图像分解为金字塔图像序列,然后对其中每一层图像使用谱残差法进行显著性检测,得到包含船舶目标的显著性子图;而后融合各子图得到最终显著图,对该显著图应用优化阈值的分割方法得到最终的检测结果。SAR数据实验结果表明,该方法具有复杂度低、检测精度高等特点,且极大降低了对先验知识的依赖。
Ship detection is an important direction of SAR image application in maritime surveillance. A multi-scale optimization threshold saliency detection is proposed in this study, for detecting ship targets of SAR image.The SAR image is first decomposed into a pyramid image sequence. Then the saliency detection is performed by using the spectral residual method for each layer in the sequence, and the salient subgraphs that contain ship targets are obtained. Finally, the subgraphs are fused and the optimization threshold segmentation method that applies to the saliency map is used to produce the final result. Experimental results show that the proposed approach has better detection performance, and it has low complexity and high detection accuracy and greatly reduces the dependence on prior knowledge.
作者
闫成章
刘畅
YAN Chengzhang;LIU Chang(Insititute of Electrics,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《中国科学院大学学报(中英文)》
CSCD
北大核心
2019年第3期401-409,共9页
Journal of University of Chinese Academy of Sciences
基金
国家重点研发计划项目(2017YFB0503001)资助