Underground Multi-Target Recognition of Ground Penetrating Radar Based on Multi-Feature Information Fusion

A multi-parameter feature and recognition method is established for GPR underground targets based on multi-feature information fusion ideas specific to complexity and diversity of detecting environment and underground media as well as non-stationarity and aperiodicity of GPR echo signals. This method carries out multi-parameter feature fusion by selecting power spectrum, wavelet packet energy spectrum and higher-order spectrum, and on this basis, executes recognition and classification with the wavelet neural network. The experimental results show that compared with single–parameter feature recognition methods, this method can effectively realize recognition of the tubular bodies, spheroids as well as geologic bodies made of metal, plastic, cement and other materials.