Segmentation of cold seep plume based on Gabor features enhancement

Accurate identifying of bubble plume is an important way to discover the outlets of submarine cold seep. The submarine cold seep plume in image features high brightness, unfixed texture, consistent direction, different scale, and unclear edge with the background environment. However, there are a large number of sediments and deep-sea dark biota that depend on gas hydrate near the submarine cold seep overflow outlet, and these complex and disordered background environments bring challenges to plume segmentation. Based on the in-depth analysis of the plume flow and background environmental characteristics, we proposed a method of the segmentation of cold seep plume based on the Gabor features enhancement. In this method, original RGB images are converted into HSV images, and the V-channel images are used to perform the Butterworth high-pass filtering, histogram enhancement, and closed-operation operations on the research object in turn to obtain the plume feature enhancement images. A Gabor filter was designed to extract the texture features of the enhanced images, and the feature vectors of the bubble plume flow are constructed by fusing the texture features and spatial features. Principal component analysis was used to reduce the dimensionality of the eigenvector, and the segmentation of the plume flow was finally realized by K-means cluster analysis. Experiments showed that compared with the Gabor and the optimal threshold segmentation algorithm, the accuracy of the proposed method was increased by 11.03% and 16.84%, and the F1 value was increased by 5.83% and 9.13%, respectively. Therefore, by highlighting the characteristics of plume flow and weakening the environmental impact of the seabed, the plume segmentation of different scales, directions, and intensities can be realized to effectively support the exploration and research tasks of submarine cold seeps.