This paper describes two recent developments at the German Remote Sensing Data Centre, which aim at the automatic generation of tie points for optical sensors and radar images, and the operational utilization of automatic tiepointing for ERS-1SAR data. In the first paper, a concept for the multitemporal registration of satellite images from optical sensors is described. This approach aims at the automated extraction of corresponding points in the images. It compresses the detection of distinct points by a so-called interest operator in one image. By cross correlation the corresponding points in the other images are traces. Subpixel accuracy is achieved by a least squares technique for digital image matching. Practical results are shown. where multitemporal image data from the Landsat TM sensor are used. The results are promising; improvements of the procedure are discussed. The second part of the paper describes the operation utilization of an automatic tie pointing, which has been realized for the geocoding of the ERS-1SAR data at the German Processing and Archiving Facility. significant features are extracted from reference data sets. The corresponding SAR image chips are stored in a database. These chips are correlated with the actual SAR data set. The procedure is describes as well as the accuracy achieved. For the time being, the determination of the reference chips is performed interactively. an outlook is given concerning several automatic approaches actually being investigated.
展开▼
