ISAR Imaging Based On Compressed Sensing

With the increasingly widespread application value of radar imaging technology in military and civilian fields,high resolution synthetic aperture radar imaging techniques and inverse synthetic aperture radar imaging techniques(SAR/ISAR) are gaining more and more concern and attention.In ISAR imaging,high range resolution can be got by transmitting a wide band signal,and then utilize the relative motion between the target and radar to format a synthetic aperture for improving the azimuth resolution. However,traditional ISAR imaging need much more measurement times for acquiring long CPI to ensure high azimuth resolution,but this is hard to meet the actual needs for target recognition and real-time monitoring.In addition,ISAR targets always are instantaneous, noncooperation and mutative,then there is a strong Doppler time-varying in the echo.So there is a very big restriction in imaging results for the long time measurement time of traditional imaging methods.In order to break the limitation of long CPI for imaging,several scientific agencies propose exploiting the compress sensing theory,which is a central issue in these years, to ISAR imaging and constructing a ISAR imaging frame based on compressed sensing(CS).In terms of the imaging results,the CS-based ISAR imaging is better than traditional methods,but imaging efficiency is still need to be improved,and only considered the local sparsely of the signal did not take full advantages of some inherent features of signal itself.In this article,for above issues,we take full account of the intrinsic properties of ISAR data and use the CS theory for super resolution imaging of initial echo data.This not only improved the imaging results and also reduced the false imaging while saving more signal energy.The main contents for this paper can be summarized as about three parts:1.A CS-based ISAR imaging combining graph cut segmentation for a new weighted method is proposed. In previously,using the same weights in the whole image can not preserve the target signal energy and suppress the background clutter very good, simultaneously.To ISAR echo,only when the neighbor pixels of a strong scatter are also strong scatters,this pixel will be consider as a real target point,otherwise,it is a clutter point.Usually,strong scatters are located in some certain areas of the imaging plane, weak scatters are located in other areas.So we can apply graph cut segmentation to divide the image into signal and background region to get the target prior region,then apply different weight methods to this two region,this would save signal energy and suppress clutter energy well when imaging.Experiments with the real data show that this algorithm effectively improved the imaging quality.2.Put forward a new algorithm of ISAR imaging based on low-rank and sparse matrix decomposition.Known the background is always almost unchanging and only sparsetarget changes by ISAR imaging characteristics,but the existing techniques only think about the sparsity and ignore the feature of background.So we regard the ISAR image is the sum of sparse target and low-rank redundant background in this method,and different from the imaging before only put single sparse constrain,but add local sparsity constraint to target and low-rank to background at the same time.Specific approach is to take the ISAR imaging problem into the imaging of target and background respectively,and believe the constructed sparse part is the ISAR imaging via the step iteration resolve the imaging problem.The real data experimental results proved the superiority of the proposed algorithm.3. ISAR imaging by subtracting the background is proposed.The ISAR imaging data is acquired in the CPI accumulation,short CPI measurement data include little target information and long CPI would result to the clutter affects target signal when imaging by the clutter accumulated while target signal accumulated.So decompose the original measurement echo into target data and background data two parts,then subtract the background data from original data and then imaging.So it can reduce the influence of background information to the target signal in the imaging process, and imaging well. Finally through the real data experiment, whether from the visual or indexes of the method have been well proved.