Research And Implementation Of Adaptive IOSGO-CFAR Algorithm Based On Clutter Classification And Recognition

Constant False Alarm Rate processing is one of the most critical technologies in radar signal processing,and its purpose is to guarantee a constant false alarm rate during the detection for targets.The general CFAR processing is predicated on the clutter distribution characteristics,and different clutter distributions correspond to different CFAR processing algorithms.Even under the same clutter distribution,a CFAR algorithm is difficult to have good detection performance in different environments.Therefore,it is of theoretical significance and practical value to study the classification and recognition of clutter and adaptive constant false alarm rate algorithm in different environments.Usually,the Rayleigh clutter,Log-normal clutter,Weibull clutter and K distribution clutter is the inherent environment of radar signal detection and signal processing.In this paper,the above four kinds of clutter are simulated and modeled,and a radar clutter classification and recognition method based on convolutional neural network is proposed.The convolutional neural network is used to extract the clue data and improve the efficiency of feature extraction.According to the fixed and varied clutter parameters,the recognition rate is 100%and 98.29%respectively,and the recognition results are accurate and stable.In the process of constant false alarms,a very important task is to determine the scale factor based on the given false alarm probability.In most case,it can only get the expression of false alarm probaility vs scale factor,and the expression of scale factor vs false alarm probability is difficult or impossile to be obtained.Therefore,the BP neural network is used to approximate the expression of false alarm probaility vs scale factor,so as to determine the value of scale factor according to the given false alarm probability.Studies shows that the simulation results are very accurate and the error is10^-7 order of magnitude,when the number of network trainings is only 40,after nonlinear transformation of natural logarithm is applied to input of BP neural networks.By simulating the classical CFAR algorithm,it is found that CA-CFAR has the best detection performance in homogeneous environment.GO-CFAR control false alarm well in clutter edge environment.OS-CFAR has excellent anti-multi-target characteristics in multi-target environments.Based on the above algorithm characteristics,this paper proposes an IOSGO-CFAR algorithm,This method uses the number of interference targets that S-CFAR can tolerate to determine whether it is a homogeneous environment or a non-homogeneous environment.When the number of tolerable interference targets is set to zero,this is a homogeneous enviroment and the CA-CFAR algorithm is selected.Once the interference target is detected,this is a non-homogeneous environment and the OSGO-CFAR algorithm is selected.When CFAR is10^-4,the algorithm simulation of IOSGO-CFAR shows that IOSGO-CFAR has better detection performance than CA?GO?SO?OS and OSGO in both homogeneous and non-homogeneous environments based on Rayleigh clutter and Weibull clutter.Then,the algorithm is designed and implemented.According to the spec,the paper gives a feasible plan for hardware implementation.The core module includes a sorting module,a sum module,a threshold generation module,an algorithm control module,a unit to be tested,and a result detection module,and verifies the correctness of each module function.This paper improve the original sorting circuit,which can delete the oldest data and insert the latest value in one clock cycle,which meets the requirements of real-time false alarm detection.In the functional verification phase,a joint simulation platform based on MATLAB and Questasim was built.The target echo signal generated by MATLAB superimposed with noise and clutter,which is input as a test stimulus into the DUV.The results of Questasim and MATLAB are compared to prove the correctness of the hardware implementation.In order to prevent the contingency of the simulation,1000 sets of test vectors that meet the test requirements are randomly generated.The simulation results show that the targets set in each test vector are dected.Finally,the verification of FPGA is carried out and the clock frequency can reach 154.967 MHz.the special signal to be observed is capture by Chipscope,and the result is correct,which verifies the correctness of the board-level function.The algorithm simulation and design implementation show that IOSGO-CFAR algorithm has constant false alarm characteristics in Rayleigh clutter and Weibull clutter environment and it shows good adaptability.