The problem of detection and localisation of ground moving targets with airborne radar is generally solved by using STAP over a short time interval leading to detection with poor angular resolution. On the other hand, SAR aims at imaging non moving objects with a high resolution which enables target recognition. For the past ten years, techniques have been studied to unify the two approaches -the ultimate objective being the detection, location and recognition of moving targets. In this paper, we present an adaptive processing to detect moving targets on multi-channel high-resolution airborne SAR images. We show how to re-localize moving targets on high resolution SAR images when only two channels are available and present results obtained on a real experiment using the RAMSES NG system.