Marine oil film identification based on GLOH, K-Means and adaptive threshold

The occurrence of marine oil spill presents a significant threat to marine ecosystems, necessitating the development of effective technologies for the oil spill identification. This study proposed an advanced recognition methodology for oil films, which integrated the Gradient Location Orientation Histogram (GLOH) feature, the K-Means clustering algorithm, and an iterative adaptive thresholding technique. Initially, the coordinate system of the original oil spill imagery was transformed to Cartesian coordinate system. Subsequently, the Laplace operator and Otsu thresholding were employed to extract co-frequency interference, with non-noise pixels adjacent to the horizontal axis for replacing the noise pixels. Afterwards, a double thresholding approach and a median filter were applied to identify and mitigate speckle noise. Upon completion of noise reduction, a gray correction matrix and local contrast enhancement technique were implemented to improve the overall visibility of the oil slick features. Thereafter, an enhanced GLOH feature extraction method, paired with the K-Means clustering algorithm, was utilized to delineate the effective wave echo regions. Finally, the iterative adaptive thresholding method was applied to achieve precise and efficient segmentation of the oil spill regions. This proposed methodology demonstrated the capability to accurately and effectively detect oil spills, thereby offering scientific and technological solutions for oil spill emergency response efforts.