Feature Extraction and Recognition Based on SVM

As a mature detection technique, ground penetrating radar (GPR) is applied into many fields. The GPR signal explanation and recognition is so important that it affects the result reliability and accuracy of the detection. The support vector machines can obtain the overall optimal solution in sample less situations. It has solved the inevitable partial minimum problem and overcome the disadvantage, which the traditional neural network cannot avoid. In this paper the GPR signal explanation model is established based on the support vector machine and the dyadic wavelet transform (DyWT) theory. It is applied in the counterfort of railway disease detection. The experiment result proved the method is valid, and it can enhance GPR explanation precision and efficiency. The recognition ratio can reach 91.2%.