Algorithm Of Synthetic Aperture Radar Image Formation

Synthetic aperture radar (SAR) is an active microwave sensor, which can produce high resolution images of the earth's surface, avoid light and weather related limitations and operate any time of day. These distinguished properties have led to SAR widely and increased use in automatic target recognition, object surveillance and related fields. So it is of great theoretical and application value to research the synthetic aperture radar's image reconstruction algorithms. With regards to image reconstruction algorithm, the resolution of conventional methods is limited by system's achievable bandwidth and they suffer from noise and side lobe interference, which leads to poor performance. So we look for a new image formation algorithm with better performance.In this paper, we first present the relationships among reflectance density function, echo signal and projection of synthetic aperture radar. Based on this point of view, we propose a tomographic SAR image formation model. Also, we describe the factors which affect the range and azimuth resolution of SAR. To overcome the challenges introduced by the complex valued reflectivity and random phrase, we present a non-quadratic regularization method. This is achieved by incorporating prior information about the behavior of the field and other interested features into image formation through the minimization of an objective function. This objective function includes three parts, representing the difference between true value and estimated value, point feature and region feature respectively, which complicate the numerical computation. To provide an efficient and robust solution, we introduce quasi-Newton method and Hessian matrix into computation by extending ideas from non-quadratic regularization. Also, this paper analyzes the selection of regularization parameters and gives the reasonable scope for the parameter selection. The experimental results show that our proposed method can highlight the point features as well as region features, has high resolution and good performance of noise and side lobe suppression and is in favor of target recognition.