Marine oil film detection method based on growing hierarchical neural gas network and multi-scale threshold segmentation.

The increasement of offshore oil extraction and transportation has brought increasingly severe oil spill risks, seriously endangering the marine ecological environment. Therefore, it is urgent to develop advanced and reliable oil film monitoring technology. An oil film detection method is proposed based on the Growing Hierarchical Neural Gas Network (GHNG) and multi-scale threshold segmentation to address the problem of weak oil film region features and severe environmental noise interference in marine radar images. This method first utilizes the unsupervised learning characteristics of GHNG network to perform dynamic topology learning and hierarchical clustering on preprocessed radar images, effectively distinguishing oil film regions from background interference. Subsequently, a multi-scale adaptive threshold segmentation technique was adopted to accurately extract oil film targets by fusing local thresholds from different scales and neighborhoods. Finally, the final segmentation result is obtained through noise filtering and coordinate transformation. The experimental results indicate that the proposed method provides an effective technical solution for automatic and precise monitoring of marine oil films under complex sea conditions.