基于霍夫直线检测和区域生长的合成孔径声呐图像线目标提取

Line target extraction of synthetic aperture sonar image based on Hough line detection and region growing

本文针对合成孔径声呐图像线目标提取过程中,由于线目标宽度(直径)小且图像受相干斑噪声影响严重而导致的提取困难的问题,提出了基于霍夫直线检测和区域生长的线目标提取算法。首先利用Otsu算法计算区域生长终止阈值,并对噪声进行初步抑制;然后在对图像进行霍夫直线粗检测的基础上,对检测结果进行不均匀分割,再进行细检测,得到区域生长种子点;最后利用区域生长对线目标进行还原。试验结果表明,该算法能准确地提取线目标;与传统区域生长对比,该算法选取的种子点数量提升了5.37倍,且线目标提取结果具有更好的连续性和完整性。

A line target extraction algorithm based on Hough line detection and region growing is proposed to solve the problem of small width(diameter)of line target and serious influence of speckle noise on the image in the process of synthetic aperture sonar image line target extraction.Firstly,Otsu algorithm is used to calculate the growth termination threshold of the region and preliminarily suppress the noise;Secondly,based on the Hough line roughness detection of the image,the detection results are divided unevenly,and then the fine detection is carried out to obtain the seed points growing in the region;Finally,region growing is used to restore the line target.Experimental results show that the algorithm can accurately extract line targets,and the number of seed points selected by the algorithm is increased by 5.37 times compared with the traditional region growth,and the line target extraction results have better continuity and integrity.