Lightning and nuclear explosion pattern recognition from optical and electromagnetic data

Lightning electromagnetic and optical pulse data in 2009 and 2010 Hangzhou Zhejiang province and historical data of nuclear explosion were used in pattern classification research. The data preprocessing includes interpolation, filtration, and normalization so on. The wavelet coefficient energy entropy, wavelet pack energy spectrum, and wavelet box counting dimension was introduced to extract feature of the nuclear and lightning electromagnetic pulse data. And the energy spectrum percentage extraction feature method was used in optical ones. Cross validation method was used to optimum the parameter of the Support Vector Machine model. Eigenvectors were fused and more satisfaction results were gotten.