A HYBRID SEABED CLASSIFICATION METHOD USING AIRBORNE LASER BATHYMETRIC DATA

In recent years, Airborne Bathymetric Light Detection and Ranging (LiDAR) has been applied intensively to map coastal depth as well as for seabed classification. In this study, we proposed a hybrid K-means and Support Vector Machine (KSVM) algorithm based on depth-derived gray-level co-occurrence matrices (GLCM) from bathymetric LiDAR. First, the calculated GLCM data set was used to sort K-means into various clusters. Second, training samples were selected on merged clusters before applying SVM classification. Finally, we evaluated the proposed hybrid algorithm in overall accuracy and the Kappa index. Compared to pure SVM, the proposed hybrid KSVM improved the overall accuracy by 24%, and the Kappa index by 0.31. The results showed that the proposed KSVM method provided promising results, in terms of accuracy and visual inspection. The benefits of the proposed classification method applied unsupervised classification of K-means as prior information for unseen seabed sediment types. This method was useful, particularly when only depth-derived information was available, or where the intensity/waveform had poor discrimination properties.