V.V. Glazkov

The article deals with building of identification device of motorized objects wheeled and tracked vehicles in the radio. To solve this problem, selected 3-millimeter range, as it allows you to create a narrower pattern and identify new informative features. Informative features identified on the basis the mathematical model of radar input signals. This model is based on the use of multi-point model of radar targets. The mathematical model is obtained for the two types of targets: truck ZIL-131 and tank T-72. As an informative feature invited to consider: counts of the signal power spectral density, the relative width of the signal and the signal strength. To ensure the invariance of the informative features to velocity of the object and the and distance to it, is their normalization. Methods of discriminant analysis shows that the most informative feature is the counts of the normalized power spectral density of the signal. Construction of an optimal structure of the identification device is difficult, so the article proposes to transfer to quasioptimal algorithms. Made reducing the vector of informative features dimension by principal component and build identification device based on neural network. Selection of an optimal neural network structure is carried out experimentally. Proposed identification device using digital signal processing. Made the functional diagram of the identification device.