Pattern recognition of multiple signals from ground penetrating radar for metal and plastic objects

The author describes methods for pattern recognition of multiple time signals of embedded targets in water or soil. The received time signals are investigated using characteristic features of metal and plastic pipes in water for better detection of plastic and for soil with noisy signals. The classification of time signals is done by statistical signal parameters, ARMAX, Prony modelling, a new difference vector method of the model transfer function H(z) in the coefficient vector feature space, principal component analysis (PCA), and the classification (more than the estimation or identification) of radar targets via complex frequencies (CNR - complex natural resonances) of these multiple time signals.