Ground Moving Target Indication In SAR Images

Ground moving target indication (GMTI) using Synthetic ApertureRadar (SAR) has been investigated intensively in many civilian andmilitary applications. The key point of SAR-GMTI is to eliminate groundclutter. As SAR imaging area is getting larger and larger, and the resolutionis getting higher and higher, improve the computation efficiency is veryimportant. In this paper, we investigated GMTI using single-channel SAR.We derived imaging and Doppler spectrum of moving targets andstationary targets, and proposed two new algorithms based on localtwo-look similarity and based on Doppler spectrum analysis respectively.In this paper, we analyzed the difference in imaging between movingtargets and stationary targets. In the two looks of the scene, a stationarytarget has the same positions, but a moving target has different positions.Therefore, moving targets can be indicated by the patch-by-patchsimilarity detection between the two looks. In the similarity detection, foreach patch in the first look, only the normalized correlations with threepatches in the second look are calculated, so this algorithm iscomputationally efficient. When compute the local two-look similarity, weuse normalized correlation. We select the threshold according to thestatistic of above calculated quantities, so the threshold is adaptive to theSAR image. The algorithm is computationally efficient, and it can indicatemoving targets accurately.Since moving targets cause variations of Doppler shift and variationof Doppler bandwidth, we can also indicate the moving targets byanalyzing Doppler spectrum. We also process the SAR imagepatch-by-patch. For each patch, after clutter locking one can find the Doppler center. And the Doppler bandwidth of the patch will also becalculated. Get statistics of above quantities, and use them to construct ameasure. Select a threshold, and then areas of moving targets will be kept,while stationary targets will be eliminated.