Research On Underwater Target Recognition Technique Based On Support Vector Machine

Statistical Learning Theory or SLT is a small-sample statistics by Vapnik et al, which concerns mainly the statistic principles when samples are limited. SLT provides a new framework for the small sample statistical pattern recognition and the general learning problem. Base on this theory, a new method for pattern recognition —Support Vector Machine or SVM is provided, which can solve small sample pattern recognition problems better. According to the analysis of the actuality and existent problems of underwater target recognition, this thesis mainly studies the underwater target recognition using SVM. Classification experiment for real underwater targets has been done and gains satisfied result. The main work and innovation can be summarized as follows:1. Analysis of the basic characteristics of ship's radiated noise such as: types of ship's radiated noise source, spectrum characteristics. Studies on the ship's radiated noise feature using higher order statistics and wavelet transform.2. Feature extraction method is presented, which is based on the 1 —dimensional spectrum, 2 — dimensional spectrum and wavelet transformsignal of ship' radiated noise. The compositive feature vector of ship's radiated noise is constructed,which includes the sub-band distributingfeatures of 1 — and 2 — dimensional spectrum and scale-energy feature.3. Base on the analysis of feature optimization methods, K-L transform is applied to optimize the compositive feature vector and the experimental result indicates that K-L transform can optimize the feature selection process.4. SVM for the underwater target recognition is studied. According to the analysis of classification experiment for real ship targets, this method can be used to classify three types' ship and gains satisfied result.5. Least Square Support Vector Machine (LS-SVM), which can reduces the burden of the SVM and improves its train speed and algorithmic complication is researched and applied to classify three types' ship.The experimental results show the feasibility and validity of this method.