一种合成孔径声呐图像目标分割方法

合成孔径声呐图像的信噪比低于普通光学图像,使图像分割成为合成孔径声呐图像处理中的重要环节。本文研究了表示合成孔径声呐图像数据分布的瑞利混合模型,结合马尔科夫随机场模型,将其应用于声呐图像水下目标(亮区)分割;通过最大期望算法分别估计目标和背景的瑞利混合模型参数,并利用该参数使用Graph cut方法进行马尔科夫随机场图像分割,通过重复迭代,最后形成稳定的目标分割结果;对实际的声呐图像进行了数据分析及目标分割,结果表明瑞利混合模型在描述合成孔径声呐声图上有良好的性能,可以改善目标分割的效果。

Object segmentation method for synthetic aperture sonar images

Synthetic aperture sonar images have a lower signal noise ratio than common optical images, which makes image segmentation an important step in synthetic aperture sonar image processing. This paper studies the Rayleigh mixture model representing synthetic aperture sonar underwater image data distribution. Combining with Markov random field model, the Rayleigh mixture model is applied to the object segmentation in high resolution sonar image. The quick unsupervised iterative method is used to segment the object (highlight region). The expectation Maximization algorithm is used to estimate the Rayleigh mixture model parameters of the object and background, respectively; and the model parameters are used to conduct Markov random field image segmentation with graph cut method. Through repeat iteration, the algorithm converges, and gives stable final object segmentation result. The data analysis and object segmentation of real sonar image data were conducted in experiment; the result shows that the Rayleigh mixture model is capable of describing complex object echo distribution, thus improve the effect of object segmentation.