Research On Clutter Suppression Method Of Adaptive Neural Network For Through-the-wall Radar Imaging

With the low-frequency penetration characteristics of electromagnetic waves,through-the-wall radar technology can locate,identify and image objects behind the wall,and has a wide range of applications in military and civilian fields.In the detection,due to the influence of clutter such as walls,the propagation of electromagnetic waves is attenuated,which greatly reduces the intensity of the target echo,and brings great difficulties to the target detection and positioning and imaging.It means that the suppression of wall clutter has become the primary task,and the existing traditional clutter suppression method cannot achieve the effect of coexistence between the filtering of the wall and the removal of other noises,and there is still a situation of low signal-to-noise ratio of the target.Therefore,this paper proposes an adaptive convolutional neural network method to suppress wall clutter in radar signals and reconstruct the target by deep learning imaging method.The research content of this article is as follows:Firstly,the research background and significance of TWR technology and application are introduced,and the relevant research status at home and abroad is systematically described from three aspects: through-the-wall radar imaging,clutter suppression method and deep learning imaging.By studying the principle of TWR imaging and the propagation mechanism of electromagnetic waves,the working mechanism of ultra-wideband through-the-wall radar and electromagnetic simulation based on the gpr Max simulation platform and FDTD method are also studied,and the influence of amplitude attenuation and dispersion,propagation form and velocity of electromagnetic waves in the wall are also studied.Secondly,an adaptive neural network method based on singular value decomposition(Resvd-Net)is proposed for the wall clutter suppression problem,which changes the previous filtering idea,transforms the original radar echo signal processing problem into a target echo and clutter recognition classification problem based on deep learning,and uses threshold segmentation technology to separate the singular values of the two subspaces to separate the two echo subspaces,and then denoising the target echo to obtain a clean target echo data graph.Finally,it is confirmed that the inhibition effect of the method is higher than that of the two traditional methods,and it is further verified by the value of the target clutter ratio index.Finally,aiming at the reconstruction problem of the physical scene image of the through-the-wall radar,an imaging method based on deep learning is further proposed,and the original data compression preprocessing is trained and learned by means of the idea of compression perception,and the echo data before and after clutter suppression is reconstructed,and the imaging results with relatively high accuracy and strong reducibility are reconstructed by the network,and the reconstruction quality is better than that of several traditional imaging methods through the value of the root mean square error index.Moreover,the imaging results obtained after clutter suppression are clearer and more realistic.