| The detection of small slow target is always the focus of people’s research because it is concealed and aggressive.Frequency Modulated Continuous Wave(FMCW)radar has been used more and more widely in the detection of short range,slow and small targets because of its advantages such as low power consumption and no distance blind zone.In the detection area,there are static and moving clutters,which will seriously affect the detection of small targets.The existing MTI filter for suppressing static clutter has wide notches,which can easily lose the energy of small targets near zero frequency.However,the adaptive filter for suppressing moving clutter is computationally intensive and difficult to achieve in engineering.Furthermore,as the resolution of FMCW radar increases,the clutter distribution model deviates from the Rayleigh distribution,and the detection performance will be dramatically degraded by using traditional detection algorithms.To solve these problems,this paper improves the target detection performance from two aspects: clutter suppression and detector optimization.The specific work of this article is as follows:(1)A new multimode clutter suppressor is presented to overcome the problem that the existing static clutter suppression methods tend to overwhelm the energy of small targets near zero frequency and the computational complexity of the adaptive filter method for moving clutter suppression.Firstly,a narrow-notch filter is designed by using the eigenvector method to reduce the energy loss of the target near zero frequency.Then,based on the possible Doppler channels of rain clutter,a Quasi-adaptive filter bank is formed by aligning the notches of two-pulse cancellers with the Doppler channels.Its optimal filter coefficients are determined by comparing the output power of each filter.Instead of estimating the clutter spectral center and the clutter spectral width in real time,the computation of the adaptive filter algorithm is reduced.The data processing results show that the ground clutter improvement factor is 4 d B higher than the traditional clutter suppression method,and the Quasi-adaptive filter bank runtime is reduced to about 0.3 times of the adaptive filter algorithm.(2)After clutter suppression,the target echo energy is still low and the remaining clutter and noise do not satisfy the Rayleigh distribution.Using traditional Constant False Alarm Rate(CFAR)detector will greatly reduce the performance of target detection.For this purpose,the target echo energy is accumulated from the measured data after clutter suppression using Moving Target Detection(MTD),then the simultaneously obtained Range Doppler Maps(RDM)are divided into blocks,and the amplitude distribution of each block’s range Doppler resolution unit is fitted to obtain the optimal distribution model of clutter and its shape parameters.Then,the shape parameter is used as a priori information to correct the label threshold factor of the CFAR detector.The false alarm probabilities of the detectors are consistent under different clutter parameters.The simulated and measured data processing results show that the Transformed Clutter Map CFAR detector(TCM-CFAR)has the best detection performance.(3)An FMCW radar target detection system based on multi-core DSP is designed on the basis of the previous research.The system is composed of a PC and a high performance So C 66AK2H12 evaluation board(EVMK2H)carrying TI.The algorithm is implemented on a multi-core DSP.First the PC imports the data into the DSP through XDS200 and the calculation is completed by the DSP.Then the processing result of the DSP is returned to the PC through XDS200,and then the visualization is realized by the MATLAB programming.In the software design process of DSP,using register optimization and single-core parallelization methods such as Single Instruction Multiple Data and software pipelining,the algorithm speed is about 13 times faster than before optimization.The algorithm is multicore parallelized using Open MP architecture.The optimized processing time is 6.45 times shorter than that of single-core parallelization.The whole computing process can be completed in 2.5ms.Finally,the processing result of the system is compared with that of MATLAB,which verifies the correctness of the DSP software optimization and the engineering value of the algorithm. |
