Synthetic Aperture Radar Imaging And Application Research

The primary applications of Synthetic Aperture Radar(SAR)are in military and civilian domains,as it allows real-time remote sensing with strong penetration,high resolution,and all-weather capabilities.The importance of SAR imaging and SAR image target recognition is therefore increasing.SAR is a special reconnaissance method,which is mainly used in the military field for detecting and recognizing designated targets and promoting modernization and informationization of the battlefield.With the rapid development of semiconductor synthetic material technology and the continuous emergence of low-power and highprecision millimeter-wave chips,there is an increasing demand for millimeter-wave SARrelated products,such as security checks and vehicle-mounted radar.Therefore,it is of great importance to improve SAR imaging technology,make it faster and more convenient,and achieve target recognition and classification of SAR images.This paper builds a system using MATLAB software,in which users can perform calibration,data acquisition,and imaging of the guide rail,making it highly convenient for users to operate.Meanwhile,in order to improve the effectiveness of SAR image target recognition,this paper utilizes the feature extraction and characterization capabilities of deep learning,and conducts experiments on SAR image target recognition based on deep learning with the following research contents:(1)In view of the complexity of SAR data acquisition,processing,and imaging operations,a hard platform was built using the AWR1843 radar and a two-dimensional guide rail.A comprehensive system based on MATLAB software platform was developed,which can visualize and facilitate the above steps,and ultimately obtain two-dimensional imaging and three-dimensional SAR imaging using multiple imaging algorithms.The effect of waveform parameters and the spatial position of the millimeter-wave radar on imaging were studied.This system can be applied in airport security checks or vehicle-mounted millimeterwave equipment,making it easier for people to use.(2)In view of the inefficiency of Res Net50 in SAR image target recognition,this paper proposes an improved network based on Res Net50,which introduces a 1×1 short-circuit branch and a Convolutional Block Attention Module(CBAM).This method first trains the target recognition model for ten military targets in the MSTAR dataset under standard conditions,and then uses the trained model to identify the validation set of these targets,resulting in significantly improved recognition performance compared with several classical convolutional neural networks(MFCNNS,DCNN,Res Net50).