| The weak maneuvering targets represented by stealth aircraft,small unmanned aerial vehicles and space vehicles seriously degrade the detection performance of traditional radar.The signal-to-noise ratio(SNR)of radar echoes received can be effectively improved by means of coherent integration processing,so that improve the detection performance of such targets.However,in the process of coherent integration,the high velocity,acceleration and acceleration of the target cause range migration and Doppler migration,which makes the echo SNR cannot be improved effectively and the detection performance deteriorates seriously.In view of the above problems,this paper carries out technical research on the coherent integration of targets under different motion states.The main work is as follows:1.Aiming at the problem of distance migration and Doppler migration of uniformly accelerated moving targets and the problem of high computational complexity of traditional algorithms,A coherent integration algorithm based on Range Frequency Reversing Transform and Second-order Wigner-Ville Distribution(RFRT-So WVD)is proposed.Firstly,the pulse pressure echo signal is reversed along the range frequency axis,and then multiplied by the original signal to correct the linear and second-order distance migration of the target trajectory.Then,the corrected signal is extracted and the acceleration of the target is estimated by the second-order WVD algorithm.Finally,after adding the velocity phase term to the compensated signal,the inverse scale Fourier transform is used to estimate the target velocity and complete the coherent integration.Simulation and experimental data show that the algorithm can effectively achieve coherent integration of uniformly accelerated moving target echoes and estimate motion parameters.Compared with the three-dimensional Fourier transform algorithm,the computational complexity is reduced by about10~3.The detection probability is better than 90%when the SNR is better than-1d B.2.For the coherent integration of targets with second-order distance migration in variable acceleration motion,geometric-auxiliary location rotation transform and periodic scaled generalized high-order are proposed ambiguity function(GLRT-PSGHAF)coherent integration Algorithm.Firstly,the concrete implementation process of axis rotation transformation algorithm in frequency domain is derived,which can effectively reduce the error in rotation process.Then,an axis rotation transformation algorithm based on geometric relation was proposed,which converted the search rotation Angle into geometric relation to solve the problem,greatly reduced the computational complexity,and realized the correction of second-order distance migration.Secondly,the estimation range of the acceleration is extended by using the idea of periodic extension of the discrete Fourier transform,and the results of the estimation parameters are avoided from exceeding the original limit range of the algorithm under the condition of not changing the basic system parameters of the radar.The generalized periodic scale high-order fuzzy function can estimate the target acceleration and acceleration effectively.Finally,after obtaining the target acceleration and acceleration estimation,the compensation of doppler migration is completed,and the coherent integration of target echo signal energy is realized.Simulation and experimental data show that the algorithm can effectively realize the coherent integration and parameter estimation of the variable acceleration target with second-order distance migration,and the computational complexity is reduced by 10~9 compared with that of the generalized radon Fourier transform(GRFT).The detection probability is better than 90%when the SNR is better than-5d B.3.For the problem of coherent integration of variable acceleration targets with third-order distance migration,product scaling periodic modified Lv’s distribution(PSPMLVD)and inverse scaling periodic discrete Fourier transform(ISPDFT)were proposed.This algorithm directly aims at the two-dimensional radar echo data after pulse pressure,and introduces the scale-period discrete Fourier transform and product processing on the basis of the improved Lv’s distribution algorithm,which has the ability of two-dimensional radar signal processing.Firstly,the scale-period-modified Lv’s distribution algorithm extends the range of acceleration at a single range frequency.Then,by multiplying the results of all the distance frequencies,the spectral peaks after period extension are suppressed and the spectral peaks directed by the real parameters are accumulated,and the estimated accurate acceleration and acceleration after extension are obtained.Finally,the inverse scale periodic discrete Fourier transform is used to process the compensated and accelerated signals,estimate the target velocity and complete the coherent integration.The algorithm has good cross-term suppression effect,can adapt to all variable acceleration motion models,and can change the extension period according to the actual situation to expand the range of parameter estimation.Simulation and experimental data show that the algorithm does not need model approximation and parameter search,and can freely adjust the size of parameter estimation range.Compared with GRFT algorithm,the computational complexity is reduced by about 10~8,and the detection probability is better than 90%when the SNR is better than-3d B.4.According to the integration problem of coherent integration and target detection,a method based on Residual Network(Res Net)is proposed to accelerate the calculation speed of coherent integration and realize target detection after integration.Firstly,a generalized ISPDFT(GISPDFT)algorithm was proposed on the basis of the ISPDFT algorithm proposed in this paper.Secondly,to solve the problem that the algorithm needs to search acceleration parameters and expansion multiples,a Res Net-50 network is proposed to predict the rough estimation range of search parameters and expansion multiples,so as to reduce the number of search accelerations and accurately estimate the expansion multiples.Finally,a target detector based on Res Net-34 is designed.In order to extract more feature information,the 7×7 large convolution kernel in the network is designed to be replaced by three small 3×3 convolution kernel.The simulation and experimental data show that compared with GRFT,the computational complexity of Res Net coherent integration algorithm is reduced by about 10~3,and the anti-noise loss is about 2d B.When the constant false alarm rate(CFAR)detection of Res Net target detector is higher than 90%,the required SNR improvement is about 1d B. |
PDF Full Text Request