吕新荣, 李云, 温春苗, 任鹏
中国石油大学(华东) 海洋与空间信息学院, 山东 青岛 266580

摘要:
为了提高合成孔径雷达（synthetic aperture radar，SAR）图像中海上溢油的检测精度，设计了一种图像显著性和均值偏移分割相结合的暗斑检测方法，验证了该方法在SAR溢油图像暗斑识别中的有效性，为溢油区域的准确检测奠定了基础。针对SAR溢油图像中暗斑区域在人类视觉中属于感兴趣区域的特性，利用图像显著性检测方法对整幅SAR图像进行处理生成显著性图像，然后利用均值偏移方法在SAR图像的Lab空间将图像分割成不同区域，最后在显著性图像上利用每个区域的平均显著性值与动态阈值之间的关系确定溢油图像中的暗斑区域。实验结果表明，该方法不需要人工交互，且能够有效地提取出SAR图像溢油中的暗斑区域。
关键词:合成孔径雷达(SAR)图像显著性均值偏移暗斑检测
DOI：10.11759/hykx20200324004
分类号:P76
基金项目:国家自然科学基金（61971444）；山东省自然科学基金（ZR2019MF019，ZR2016DL11）

Dark spot detection in SAR oil-spilled images based on saliency analysis
LYU Xin-rong, LI Yun, WEN Chun-miao, REN Peng
College of Oceanography and Space Informatics, China University of Petroleum(East China), Qingdao 266580, China
Abstract:
To improve the detection accuracy of sea oil spills in synthetic aperture radar (SAR) images, a dark spot detection method combining image saliency and mean shift segmentation is designed. Its effectiveness was verified in the identification of dark spots in SAR oil-spilled images, laying a foundation for the accurate detection of oil-spilled areas. In view of the characteristics of the dark spot in the SAR oil-spilled image that is relevant to human vision, the entire SAR image is processed by the image saliency detection method to generate a saliency image. The mean shift method is used in the Lab space of the SAR image to divide the image into different regions. Finally, the relationship between the average saliency value of each region and the dynamic threshold is used to determine the dark spot region in the oil spill image. Experimental results show that this method does not require human interaction and can effectively extract dark spots in SAR oil-spilled images.
Key words:synthetic aperture radar (SAR)image saliencymean shiftdark spot detection





PDF全文下载地址:

点我下载PDF