一种基于多元特征的海上目标分割方法

A MARINETIME TARGET SEGMENTATION METHOD BASED ON MULTIPLE FEATURES

提出一种结合多元特征的海上码头和船舶分割方法。由于不同种类的图像各自特点不同,多光谱图像颜色信息丰富,全色图像分辨率较高,SAR图像散射特征明显,尤其对于人造目标在SAR图像中通常表现为高亮点。利用多光谱、全色、SAR图像,提取多光谱图像的颜色特征分割水陆,然后在全色图像中提取码头的上下文特征和船舶的几何特征,在SAR图像中提取灰度统计特征,最后,结合多元特征实现海上码头和船舶的分割。实验结果表明该方法可以有效地分割码头和船舶。

An offshore pier and ship segmentation method in combination with multiple features is proposed. Since different images have different features, the multi-spectral images are rich in co｜our information, panchromatic images have higher resolution, and SAR images are significant in scattering characteristic. Especially, for the man-made objects, in SAR images they usually appear as highlighting points. We use these three kinds of images to extract the colour features in multi-spectral images respectively to segment the land and the water, and then extract the context feature of the pier and the geometric feature of the ship in panchromatic images, and extract statistical intensity features in SAR images. Finally, the segmentation of offshore pier and ship in combination with multiple features is realised. Experimental results show that the method can effectively segment the pier and the ship.