.png){kind=icon}
.png){kind=icon}
Return
An Automatic Data-Driven Method for SAR Image Segmentation in Sea Surface Analysis
.png){kind=icon}
PRE
AI
DOI：10.1109/TGRS.2017.2769710.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In the context of synthetic aperture radar (SAR) image segmentation, this paper proposes a new automatic unsupervised method addressing sea surface analysis with a focus on oil spill and ship segmentation. Being an evolution of an existing algorithm originally devoted to the detection of a single region of interest, the present method performs a global image segmentation of the whole image. The processing is independent of any model and is driven by data informative content along with intermediate results. Based on graph theory, it makes use of a new defined cost function and assigns cost values to the vertices rather than to the edges of the graph. The experimental results achieved and numerically evaluated on synthetic and real SAR images prove that the method is robust and repeatable, and it does not involve restrictions on image modality acquisition or sensors and does not require radiometric calibration. It can work on amplitude or intensity SAR images, independently on frequency band, polarimetry, and spatial resolution. Qualitative and quantitative performance analyses are carried out along with a comparison with other published works in the same application. Good results are achieved for both oil spill and ship segmentation and robustness by changing seed points position and number. Errors exhibit stable behavior when increasing the number of seed points. Finally, in contrast to most of the existing methods, the proposed technique does not depend on parameters and is generally more robust.
Keywords:
Automatic graph-based segmentation
multi-seed
sea monitoring
synthetic aperture radar (SAR) images
unsupervised
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
8.6
Papers:
2.1W
Citations:
10.7W