A Sea–Sky Line Detection Method Based on the RANSAC Algorithm in the Background of Infrared Sea–Land–Sky Images

Abstract

Subgrade sea and sky monitoring equipment requires the accurate detection of threat targets in a given area. Due to the extremely complex sea–land–sky backgrounds, the sea–sky line is often submerged in the background. Therefore, we propose an algorithm for accurately detecting sea–sky lines under a complex sea–land–sky background. Based on the analysis of infrared images with sea–land–sky backgrounds, we segment images using the k-means algorithm. To use the random sampling consistency (RANSAC) algorithm to fit the sea–sky line better, we divide the images into nonuniform segments and count the row mean-value gradient trough. The experimental results show that the sea–sky line can be detected in 1215 out of 1227 pictures. The test success rate is 99%, and the difference from the actual sea–sky line is less than 3 pixels. The method presented has higher adaptability under a subgrade sea and sky monitoring environment.