Target Passive Detection Technology Based On Cooperation Of Multiple Types Of External Radiation Sources

Passive radar positioning systems based on external radiation sources do not emit signals to the outside world,but use signals emitted by third-party radiation sources for detection.The signals emitted by external radiation sources are received by the receiving station after being reflected by the target,and the receiving station estimates the echo Parameters in order to detect the target.The external radiation source radar detection system has strong concealment and anti-jamming capabilities,and also has strong detection capabilities for stealth targets.The external radiation source radar system uses the signals emitted by external radiation sources such as FM FM radio,GSM base stations,GPS satellites,and terrestrial television broadcasts to detect the target.These radiation sources are densely distributed in inland areas.If these different kinds of external radiation source signals can be comprehensively used,a reasonable multi-platform external radiation source cooperative detection system can be designed,compared with a detection system that relies on only one external radiation source,the stability and reliability of multi-platform external radiation source radar system detection will be greatly improved.In this thesis,by using multiple types of civil radiation sources as external radiation sources,a passive detection method based on multi-platform radiation source coordination is proposed.The main work and innovations of this thesis are as follows:(1)This thesis proposes an AE-CNN network framework based on the self-encoder network and convolutional neural network to solve the problem that traditional DOA estimation algorithms are greatly affected by array errors under the background of Gaussian noise problem.In view of the above problems,the neural network framework includes two parts,a linear classifier and a convolutional neural network(CNN).The linear classifier mainly plays the role of preliminary classification.The output units of the linear classification network are divided into P groups.Each group of output units represents a spatial sub-region of a certain angular range.The signals are processed by the linear classifier and output in different output units,which represent that the signals come from the spatial sub-region represented by the output unit,and then the convolutional neural network is responsible for accurate DOA estimation in each sub-region.This method has high adaptability to antenna array defects,and can be more commonly used in unknown scenarios,and has better generalization ability than traditional methods.(2)Aiming at the problem that the target detection of multi-platform external radiation source cooperation under the background of Gaussian noise is greatly affected by noise and the estimation accuracy is not high,this thesis first proposes a Linear Canonical Transformation based Cross Ambiguity function(LCTCAF)method based on linear regular transformation.The method can estimate the distance from the target to the receiving station and the speed of the target.In addition,basing on the mapping relationship between target distance,radial velocity and time delay and Doppler,this thesis proposes an Reverse transformation based Cross Ambiguity Function(RECAF).Then the estimation results of the two methods are converted into estimates of the target's position and velocity.This thesis verifies the effectiveness of the two methods through simulation experiments,and proves that the RECAF method suppresses noise,and has better estimation performance than the CAF method and the LCTCAF method.