Research Of SAR Countermeasures Based On Shadow Feature

Due to its advantage of all-weather and all-time,SAR has played an irreplaceable role in military fields.With the improvement of its military function,research aiming at SAR jamming and anti-jamming becoming more and more important.SAR deception jamming seriously affects the reliability of SAR image interpretation by adding illusory targets on SAR images.As a kind of active remote sensor,shadow feature of SAR images is obvious in the side looking position.Therefore,aiming at deception jamming,this thesis researchs SAR countermeasures based on shadow feature.The main contents and innovations are as follows:Basic principle of SAR,the realization of SAR deception jamming and the shadow feature of SAR image are expounded at first.It is shown that the essence of SAR deception jamming is the superposition of jamming echo and real echo,and it is difficult to simulate the shadow feature of false target by deception jamming.So it is possible to carry out anti-jamming method by recognising shadow feature in SAR image domain.Then neural network that is used in recognition and its training method are briefly introduced.The convolution neural network model of LeNet-5 structure is improved for SAR target classification.The active functions in convolutional Layer and full-connection classification layer of the LeNet-5 structure are respectively changed into ReLU and softmax functions.Max sampling method is used in pooling layer.In order to improve recognition rate,layers and parameters are adjusted.Considering that there is no public SAR image database with deception jamming,this thesis studies the SAR imaging simulation based on electromagnetic simulation software.By calculating the target surface electromagnetic flow distribution under radar irradiation,echo simulation and imaging processing,imaging results of complex target under deceptive jamming are provided.This method simulates the sample libraries for SAR deception anti-jamming based on target recognition.Deception anti-jamming method in image domain is propose.Traditional SAR deception anti-jamming methods rely on complex modulation of transmit signal to suppress the imaging of jamming echo.However,these methods can’t deal with the case that jamming echo has already imaged.In this thesis,SAR images are classified by convolution neural network at first.At this moment,outline information of targets is obvious,different types of targets could be well distinguished.But shadow feature of real targets is too inconspicuous to be recongnised.After this,OTSU and morphological operation are combined to heighten shadow feature at the cost of losing outline information.When the images are prepossed,shadow feature can be recognized by a new network,and deception targets could be marked.Two SAR deception methods based on active shadow elimination are proposed from the aspects of scatter-wave and deception jamming.In the case of scatter-wave jamming,positions of target and its shadow are same in azimuth,and in range direction the position of shadow is delayed,which is similar with the result of scatter-wave jamming.So shadow feature can be eliminated by calculating the position of jammer and scattering region.In the case of deceptive jamming,it is known to us that deception jamming can add jamming targets to SAR image in a specific position.In this thesis,shadow region of target in SAR image is uperimposed with background by computing position of target shadow and modulating the deception jamming echo.With shadow eliminated,real target looks similar with the deceptive target what deception jamming create,and the popose of disturbing SAR image interpretation is achieved.