Angular response classification of multibeam sonar based on multi-angle interval division

摘要

In seabed classification techniques based on multibeam sonar, the classification method using angular response information of backscatter strength data has shown the potential for practical application and can be supported by the seabed high-frequency acoustic scattering theory. However, classification imagery produced by this method has a relatively low spatial resolution, which is limited to the swath width of multibeam sonar. To solve this problem, the angular response curve (ARC) of backscatter strength has been divided into several intervals, and the fragment of the ARC within each angle interval has been processed separately to extract the features within the sample data. Features extracted in this way include the average value of the fragment of the ARC, the first order derivative, the second order derivative, etc.; and a set of feature vectors has been formed in each interval. Finally, the feature vector samples in each interval are independently trained and predicted by a support vector machine classifier. For this method, the spatial resolution of the classification imagery is the size of the angle interval coverage, limited to the number of divided intervals. Therefore, the area of the region represented by each feature sample has been reduced, which can improve the spatial resolution of classification imagery. Meanwhile, the results of processing the experimental data show that this method also has relatively good classification performance.