A method for feature extraction and classification of marine radar images

Abstract

There are ship images (target) and non-ship images (noise) on radar images. In order to obtain other ship information from radar images, it is necessary to select and acquire ship images on radar images. Ship images are selected and acquired by navigation officers based on their observation skill and experience. The future purpose of this study is to automatically detect ship images on the radar. Therefore, in this paper, we propose the method for dividing ship images from non-ship images by the image processing and the cluster analysis using radar raster images. Many image feature points were extracted by the image processing using radar raster images. As a result of the cluster analysis using these image feature points, it is possible to detect about 99.8% ship images from radar raster images. However, there were many cases that non-ship images were classed as ship images. Therefore, the accuracy rate of cluster analysis results in this study was about 83%. In other words, it was possible to fairly determine about 83% images in this study.