关键词: 水边线提取, G0分布, SAR图像, 斑点噪声

Abstract:

The water line is the demarcation line between land and sea areas. It is important interaction zone for global ecosystem conservation. As a result, it is vital that the rapid, accurate, real-time extraction of water line. Remote sensing solutions using Synthetic Aperture Radar (SAR) are playing an increasingly important role in monitoring water line and have getted much more attention from scholars and related department around the world. The sensor has two types: satellite SAR and airborne SAR. Satellite SAR has many advantages, for instance wide range observation, timely data available and all weather operation. As a result, it is a most suitable sensor for monitoring water line in marine environments. At present, the commonly used satellite SAR sensors for this purpose including Seasat SAR, Jers-1 SAR, Sentinel-1 SAR, and so on. In the SAR images, water line usually shows the edge characteristic, so the edge detection based algorithms are the most commonly used in coastline extraction. The classical edge detection operators including Canny operator, Sobel operator, Roberts operator, Prewitt operator and Laplacian operator, and so on. The operators have some advantages such as simple, fast speed, but they always effected by speckle noise inherent in SAR images, as a result, they can not obtain accurate water line results. So noise is a major issue for the task of boundary detection by the SAR images. Instead of combating the noise, we use a technique for boundary detection in SAR images based on the statistical properties of speckled data. The G0 distribution is a special form of G model, its parameters are very sensitive to the surface roughness and succeeds in characterizing a wide range of areas as sea, mountain chain and urban areas in speckled images. And the parameters required for boundary detection is extracted with moment estimation. Moreover, relative to G distribution, G0 distribution does not include the complex Bessel function, and is a kind of simple statistical distribution model with less parameter. According to the estimated parameters of G0 distribution, the marine and land areas can accurately be separated. In order to extract the water linefrom SAR image, a G0 distribution based algorithm is proposed in the article. First, the domain of SAR image is divided into a set of sub-blocks with the same size, and the grayscales of pixels in each sub-block are assumed to be identical and independent G0 distribution. The roughness of each sub-block and scatter shot parameters of the distribution are obtained with moment estimation. By thresholding the roughness parameter, the rough sea area can be divided, and then its geometric center is determined. On the rays started at the geometrical center, the cut-off points between sea and land are located with likelihood function. Consequently, the water line is formed by linking all the cut-off points. To prove its feasibility, a process of water line extraction has been tested with simulated and real SAR images by the method. Qualitative and quantitative results show that the proposed method can extract water line from SAR image effectively and efficiently.

Key words: water line extraction, G0 distribution, Synthetic Aperture Radar image, speckle