CFAR Detection Of Radar Signal And Estimation Of Spatial Signal Sources

In automatic signal detection system, the false alarm rate with a fixed threshold will intensely increase when the voltage of noise or clutter raises several dB. The increase will lead to the computer overloading so that it is impossible to make a correct decision for enough SNR. Thus, for the fluctuated noise and clutter, constant-false-alarm-rate (CFAR) processing which sets the threshold adaptively makes the false alarm rate constant. That is constant false alarm rate(CFAR) detection.High resolution radar obtains wide application and consideration compared with low resolution radar in anti-clutter, precise detection, high precision tracking and target recognition and so on. Under high frequency wide-band radar, the sizes of warships or airplanes are larger than resolution cell of radar and targets echo is distributed in several range cells. A new method of CFAR detection of distributed target is proposed based on the fusion of time and frequency dual-threshold detection results. Then, according to the optimal matched filtering and principal component analysis theories, another method of CFAR detection is proposed making use of all range profiles among angle. This method has not only good detection performance but also good steadiness for range profiles of different angle.Recently, many scholars start to deal with the problems that exist when spatial spectrum estimation is applied into practice. Spatial signal sources estimation is one of the most scabrous problems. The existing estimation algorithms need to evaluate sampling covariance matrix and eigenvalue decomposition which has a great operation, and also need a great number of snapshots for estimating the covariance matrix, so that it is difficult to implement practically. A dual-threshold RELAX-CLEAN estimation algorithm is proposed based on the spatial CLEAN algorithm. The algorithm does not evaluate the covariance matrix and eigenvalue decomposition and has a good performance when few snapshots needed. However, the algorithm RELAX needs large calculation for estimating signal parameters. Aimed to the above disadvantage, another algorithm based on fixed threshold is put forward. The experiment shows it has lower estimation precision than the dual-threshold RELAX-CLEAN, but the operation is small and the implementation practically is easy.