| 摘要: |
| 针对侧扫声呐图像斑点噪声强、背景海底散射干扰严重，海底目标轮廓自动提取困难的问题，提出了一种基于K-means聚类与数学形态学相结合的海底目标轮廓自动提取算法。为克服噪声干扰，该算法首先利用中值滤波去除侧扫声呐图像中的强斑点噪声；然后采用K-means聚类算法对侧扫声呐灰度图像进行分割，并二值化，除去大部分海底背景噪声，初步提取出目标；接着利用数学形态学运算去除提取结果中的孤立噪点，并填充目标内部孔洞，得到连续化、圆滑的目标边缘；最后对处理后的侧扫声呐图像进行边缘检测，提取出目标轮廓。实验结果表明：该算法思想简单易行，具有很强的克服背景噪声的能力，自动提取的目标轮廓连续性较好，结果准确可靠。目前，在侧扫声呐图像目标轮廓提取过程中，主要采用人工方式，自动性较差，效率较低。本文算法可以实现目标轮廓的自动提取，提高效率，具有较强的实用价值。 |
| 关键词: 侧扫声呐图像 K-means聚类 数学形态学 边缘检测 自动提取 |
| DOI：10.11759/hykx20181217001 |
| 分类号:P733.23 |
| 基金项目:国家自然科学基金项目（51179169） |
|
| Automatic extracting target contour of side-scan sonar images by uniting K-means clustering with mathematical morphology |
|
WANG Tao, PAN Guo-fu, ZHANG Ji-bo
|
|
Second Institute of Oceanography, Ministry of Natural Resources, Hangzhou 310012, China
|
| Abstract: |
| To address the issues of strong speckle noise and severe scattering interference of seabed background in side-scan sonar images along with difficulty in the automatic extraction of seabed target contour, an automatic extraction algorithm of seabed target contour based on K-means clustering and mathematical morphology is proposed. First, the algorithm uses median filtering to eliminate strong speckle noise in the sonar images. Second, the K-means clustering algorithm is used for the segmentation of the side-scan sonar images, and the process of binarization is used to remove most of the seabed background noise along with a preliminary extraction of the target. Third, mathematical morphology is used to eliminate the isolated noise in the extraction result, and the internal hole of the target is filled to obtain continuous and smooth target edge. At last, edge detection is performed on the processed side-scan sonar images, and the target contour is extracted. The experimental results demonstrate that the proposed algorithm is simple and easy to implement and has a strong ability to overcome background noise. The automatic extraction of the target contour has good continuity, and the results are accurate and reliable. At present, in the process of extracting target contour of the side-scanning sonar image, the manual method is mainly adopted, and the automaticity is poor and the efficiency is low. The proposed algorithm can achieve the automatic extraction of target contours, improve their efficiency, and has a strong practical value. |
| Key words: side-scan sonar image K-means clustering mathematical morphology edge detection automatic extraction |
