基于海图信息的SAR影像海陆自动分割

海陆分割是对海洋SAR图像进行船舶检测的基本前提之一,传统的手动分割方法可以很精确,但操作繁琐,速度较慢,不适合于批量图像的处理,而海岸线检测算法抗噪能力较低,对于近海多岛屿地区检测效果较差。提出了一种基于海图信息的SAR影像海陆自动分割方法,该方法利用SAR图像的地理信息,有效地与基于地理先验信息的海洋区域矢量图层相叠加,将对SAR图像的海陆分割问题转换为对矢量图层中多边形矢量元素区域的判断,并使海陆分割问题实现了自动化。最后使用Radarsat-1数据和ALOS PALSAR数据对该方法进行了验证。实验表明,近似自动分割与手动分割效果非常接近,对船舶检测几乎没有影响,且运行速度较快,适于图像的实时处理。

Automatic SAR Sea-land Segmentation Based on Sea Chart Information

Sea-land segmentation is a basic premise of the SAR vessel detection. The tradition manual technique is precious, but it operates complex and is time-consuming, and the coastline detection algorithm is poor in noise resistance, which is not suitable for the coastal area of islands. In this paper, an automatic segmentation method is proposed, taking advantage of the geography information, overlapping SAR with shape layer of the same sea area and turning Sea-Land segmentation to the judgment of the polygon shape element in the shape layer, which operate automatically. Finally. we test the technique by Radarsat-1 and Alos-Palsar images. The automatic segmentation technique has almost no influence to the vessel detection, and the result is close to the one of the tradition manual method, besides, it runs faster, suitable for the real-time operation.