The Battlefield Reconnaissance Radar Target Recognition Technology Research

With the evolution of modern war, the battlefield surveillance radar is becoming more and more important.On the battlefield, we not only hope to detect the distribution of the targets, the higher requirements are to detecte the authenticity,enemy property, the threat level of the targets around through the radar. The research on the battlefield surveillance radar’s target classification and recognition is always a difficulty in the field of radar target recognition, the resolution of battlefield surveillance radar is generally relatively low, the feature information which is detected of the targets is limited, so it is difficult to get the information which can reflect the target property.In this article the target recognition technology of an quasi-continuous wave(QCW) battlefield surveillance radar, the main work includes echo preprocessing,target feature extraction, classifier design and identify the results of analysis.The first step is the preprocessing of the echo data, which lay the foundation of the target feature extraction. Firstly the paper discusses the signal’s form of a pseudo-random code, and then is about the digital down-conversion, matched filter, sidelobe suppression, Doppler filter, distance door rearrangement, power spectrum estimation through these steps to get the power spectrum of the radar signal echo, and finally we finish the target detection through CFAR detection algorithm, to the data pending feature extraction according to the goal test results.Then, the paper extractes the feature of the target through the spectrum waveform entropy, relative RCS, wavelet decomposition feature.The previous two extraction methods is based on the structural characteristics of the spectrum, while the wavelet decomposition is based on the echo signal pulse pressure the accumulation of data, which investigated the characteristics of different target echo in time-frequency.Finally, the paper uses the nearest neighbor classifier, support vector machine classifier, BP neural network to identify and classify the feature extraction data samples, analysises and compares the advantages and disadvantages of the three classification methods, and then combines the three classifiers with a combination of classifiers, the ultimate recognition rate is up to90%.