Nowadays,information geometry detectors mostly utilized covariance matrix model and measured the difference between the sample data from the cell under test and clutter data on the matrix manifold to distinguish them for radar target detection. However,under complex clutter backgrounds,the received signal with target echoes is clutter-domi-nated due to the low signal-to-clutter ratio,so the similarity between them in terms of statistics leaded to the unavailable distinguishability,so the performance advantage of information geometry detector was limited. To break through this lim-it,this paper proposed the information geometry detector based on a joint optimization of feature and metric. Specifically,this paper first designed the flexible framework of information geometry detector with a changeable signal feature and a metric. Then,on the basis of this framework,the Neyman-Pearson criterion based joint optimization with respect to fea-ture and metric was established. By utilizing the locally flatness hypothesis and multilayer perceptron,the decision vari-ables in the optimization problem were parameterized,and then the two-stage algorithm for this optimization problem was deduced. Based on the simulated data and real-recorded sea clutter data,the experiments show that the superiority of the proposed method than existing information geometry detectors and typical detection methods. Moreover,the experimental results demonstrate that the proposed method possesses the great advantage in slow moving target detection when the tar-get Doppler closes to the peak of the clutter spectrum.
维普
万方数据