A. G. Aganin, M. Yu. Mozgonov

It is supposed to use long time of coherent accumulation of signals in onboard radars of perspective fighters for detection range augmentation that will lead to undesirable increase in time of space survey. The application of sequential test with coherent accumulation of signals will allow to reduce out a mean detection time and to improve resolution by Doppler frequency. Onboard radar works with background of clutter from various sources. Particularly reflections from an underlying surface will be presented as well. To reduce effect of stationary (within accumulation time) clutters it is possible to apply signal weight processing. As the spectrum of such clutter is unknown a priori, adaptation is necessary. Synthesis of sequential coherent multi-channel adaptive algorithms of the detection was not considered in the existing literature, concerning the account of the changing sizes of elements of the Doppler frequency resolution, statistical influence between channels, and also weight processing. The work purpose was development of sequential coherent speed multichannel adaptive algorithm of signals detection against non-uniform clutter. As criterion of optimality the maximum of probability of correct detection (PCD) has been chosen at the set characteristics: probabilities of a false alarm (PFA) and average quantity of time intervals of acceptance of the terminal decision at the validity of zero hypothesis H0 (the target is not present). Varied parameters: thresholds and the size of time intervals on all steps of observation. The direct modeling (Monte-Carlo) was used in the base of the synthesis method. As an example, the algorithm with unique thresholds on each step on the extreme statistics was synthesized in work. Arrangement of Doppler filters with constant step was used, corresponding to the maximum time of coherent accumulation. Researches of stability of the synthesized algorithm in presence of strongly frequency non-uniform clutter (with level jumps of spectral power density) are carried out. Two means of adapting have been observed for this purpose: on the basis of an average with censoring and on order statistics. For each way of adapting two alternatives of an estimation of noise mean square deviation in an adapting window were used: without detection and with detection of background clutter jump. It was found that algorithm with adapting on the basis of order statistics (rank is 3/4 of sample size) without detection of background clutter jump is the most preferable. In the circumstances, when clutter frequency size exceed adapting window size and clutter/noise ratio is 20-40 dB, the given algorithm admits increase PFA no more than in 4 times.