Classification using a Radial Basis Function Neural Network on Side-Scan Sonar Data

Detecting and classifying mines among natural formations and man-made debris along the sea floor can be a tedious task. To reduce operator dependency, an automated computer aided detection and classification system is needed. Our proposed automated system uses a two-step process. First the images are normalized and then a supervised learning method, radial basis function neural network (RBFNN), is applied to a side-scan sonar (SSS) data set. This method is able to extrapolate beyond the training data and successfully classify mine-like objects (MLOs).

[1]  J.T. Cobb,et al.  Sonar processing for short range, very-high resolution autonomous underwater vehicle sensors , 2005, Proceedings of OCEANS 2005 MTS/IEEE.

[2]  Jie Tian,et al.  Automated detection/classification of objects in side-scan sonar imagery , 2004, 2004 International Conference on Intelligent Mechatronics and Automation, 2004. Proceedings..

[3]  Gerald J. Dobeck.  Image normalization using the serpentine forward-backward filter: application to high-resolution sonar imagery and its impact on mine detection and classification , 2005, SPIE Defense + Commercial Sensing.

[4]  J. Mark.  Introduction to radial basis function networks , 1996 .

[5]  B. Sankur,et al.  Object Detection Using Haar Feature Selection Optimization , 2006, 2006 IEEE 14th Signal Processing and Communications Applications.

[6]  Myoung An,et al.  A group filter algorithm for sea mine detection , 2005, SPIE Defense + Commercial Sensing.

[7]  Qing-Cai Chen,et al.  GA-based feature subset clustering for combination of multiple nearest neighbors classifiers , 2005, 2005 International Conference on Machine Learning and Cybernetics.

[8]  David G. Stork,et al.  Pattern classification, 2nd Edition , 2000 .

[9]  Gerald J. Dobeck,et al.  Automated detection and classification of sea mines in sonar imagery , 1997, Defense, Security, and Sensing.