Video-like MiSAR image sequence processing

The miniature SAR-system MiSAR has been developed by EADS Germany for light weight UAVs (Unmanned Aerial Vehicle) like the German Luna system. In a special imaging mode MiSAR generates SAR image sequences with approximately 1 Hz frame rate and an overlap area of more than 90% for successive frames. Thus special video processing steps are applicable to the SAR image sequence: (1) Image based coregistration of consecutive image frames. (2) Image sequence mosaicing. Compared to strip SAR mode, this application is not limited to straight linear flight paths and constant squint angles. (3) Image stabilization, so that the fixed scene object is remaining in a constant place relative to a surrounding (virtual) image frame. Similar to (1) and (2), this application uses the registration parameters estimated by the coregistration algorithm. The algorithm was initially developed to process video-imagery and was adapted to the special characteristics of MiSAR imagery. The extracted image features have been proven well suited in video and robotics applications. After feature matching, a robust estimator computes the parameters of an adjustable image-to-image transformation. Additionally, when using SAR images, an appropriate gray value scaling must be applied to adjacent images in order to get a good mosaicing and stabilization result. The algorithm is implemented on a standard PC, is linked to the SAR processing unit, processes the image sequence in real time, and displays it for the visual interpretation by photo interpreters (PI). Our experiences show pretty good results since the registration algorithm works very well, even in cases where the SAR images show some SAR specific characteristics like spatial convolutions and displacements of structures, local blurring or changes in flight and squint direction.