基于冰情图的极地SAR海冰图像分割

Polar SAR Sea Ice Image Segmentation Based on Ice Chart

冰情图在极区安全航行、气候研究等方面具有重要价值,但其存在不能提供像素级的定位信息、对密集度的估计较粗略等缺陷。基于此,提出一种基于冰情图的边缘保持区域型MRF分割方法。依据冰情图从SAR图像中提取子图像,进行SRAD滤波、分水岭初始分割、区域型MRF分割,合并各子图像得到最终分割结果,实现人工解译和计算机解译的结合,得到像素级的结果,具有物体边缘定位准确、分割效率高、可并行化处理等优点。实验结果表明,该方法对极区SAR海冰图像均具有良好的分割效果。

Ice charts are recognized to be of significance in both shipping safety in polar regions and climate research,however they also have disadvantages on providing either pixel-level positions or accurate concentrations.An edge-preserving regional MRF image segmentation based on ice chart is then proposed,which extracts sub-images form accompanying Synthetic Aperture Radar（SAR） image,then does SRAD filtering,watershed initial segmentation and regional MRF segmentation one by one.It combines all the sub-images to get the final segmentation.It achieves the combination of manual interpretation and computer interpretation,gains pixel-based map,and has advantages on accurately locating object edges,efficient segmentation,parallel processing and so on.The high performance of the proposed method is demonstrated on the segmentation of SAR sea ice images of polar regions.