A Method for Angular Super-Resolution via Big Data Radar System

Introduction

In modern navigation radar platform, the multimedia big data processing system is always used in the radar receiver. In this case, the resolution of radar will be influenced by a large number of echo data. As one of the most important indexes of measuring object identification ability, radar angular resolution plays an important role in radar big data processing.

The objective is to distinguish multiple close targets, which are located in one range cell and the same beam. Previous research stated the radar angular resolution is relative to radar beam-width θ. The narrower beam-width is, the higher resolution is. But θ is limited by antenna aperture size d (Xing et al., 2008; Richards, 1988) namely

Here represents the wavelength. In theory, angular resolution will be improved by increasing antenna aperture. However, as a result of the constraint of antenna’s physical weight and effective wavelength, it is so difficult to design a desired antenna (Guan et al., 2012) and (Wang et al., 2015). Accordingly, folks have been searching suitable signal or information processing techniques to enhance radar angular resolution beyond those limitation factors.

For a long time, the problem of radar angular super-resolution is tackled by the techniques of Synthetic Aperture Radar (SAR) or Doppler Beam Sharpening (DBS) even though some blinds often exist in radar forward-looking direction on account of the constraint of Doppler information. For instance, an adaptive super-resolution approach is proposed via digital beam-forming (Fischer et al., 2012), by which multiple transmitters are applied to sensor to form a synthetic aperture, the size of which is almost twice that of physical aperture d, the size of which is almost twice that of physical aperture, so the resolution could be increased by time. Meanwhile, constant False Alarm Rate (CFAR) is used to select targets before actual super-resolution process and it reduces the computational complexity while estimate the targets angular information. The operation time needed for calculating digital beam-forming system was tremendously decreased. Tang introduces a method depended on a multi-channel L1 regularization model for sharpening the beam. It is demonstrated that the noise in radar receiver is significantly suppressed that the performance of beam sharpening is ensured while adopting an extended iterative shrinkage threshold (IST) algorithm to solve the regularization problem. Consequently, the noise leaks problem caused by channel pattern that is not satisfied with the strong prime condition in mono-pulse radar could be addressed (Tang et al., 2014). These methods mentioned above have significant effect on achieving radar angular super-resolution. Unfortunately, the resolution results are not satisfactory, especially the some close targets those are under considerable noise. It is so hard to determine the orientations precisely.

We know that when the radar is scanning a certain region, the echo signal is regarded as the convolution of the antenna pattern and the target angular information . We formulate it with an equation as follows

where the antenna pattern is denoted with antenna directivity, and n represents the noise in radar receiver. Hence, radar angular resolution is considered as a de-convolution process that is to restore x while g and h are known. Nevertheless, as shown in Eq. (2), the number of echo is less than that of target while angle interval between two targets is smaller than the beam-width. It is unable to obtain the target angular information at this time. Therefore, we need to use some special signal processing methods to distinguish these targets.