Micro Computer Based Sar Processing and Analysis System
Joji Iisaka, Takako Sakurai-Amano, W. Russell, I. Kenting · 1991
The usefulness of SAR images has been widely recognized and there many plans to launch satellites for SAR observation such as ERS-1, J-ERS-1 and RADARSAT. Many areas of SAR applications such as ice monitoring for ship routing, snow mapping for hydrology and crop growth for agriculture yield estimation, demand short data delivery by the nature of their application. SAR image analysis functions also need to be enhanced. Because of SAR image characteristics, more powerful image analysis functions are required to extract useful information from SAR data as well as conventional image analysis functions. SAR image reconstruction is computationally intensive and only a limited number of centralized institutes for SAR can afford to have SAR processing facilities. SAR processing throughput has been very smal1,and there are still many institutes which are processing SEASAT SAR data observed more than ten years ago. Some SAR application areas, such as oceanic wave analysis, require special treatment during the course of reconstruction. This can be difficult to do at existing facilities due to the necessity of interrupting standard production processing. Such users might need their own standalone SAR processing facilities at their own sites.Taking advantage of recent advances in image processing technology for Pc's, the authors are attempting to develop a microcomputer based SAR processing and analysis system. This system consists of a conventional 80386-25MHz PC with 16MB main memory, a 325 MB hard disk, and WEITEK and INTEL Math-CO-Processors. An LSI based image processing system, TOSHIBA DSFT 9506, is attached to perform FFT and range migration re-sampling. Raw SAR signal data is supplied via a 5 1/4 WORM disk or a CD-ROM disk, which have sufficient data storage capacities for raw SAR data sets. Using large amounts of real memory, much less swapping of sub-image data for corner turns is required. This system provides enhanced image analysis functions using a frame-by-frame based image computing technique called pixel swapping. The system will also provide enhanced functions for SAR image analysis using, as an example, fractal measure based segmentation of SAR images. Fractal measures provide scale independent measures which provide texture measures similar to human perception and can be useful for treating SAR images of different resolutions. Applying the pixel swapping method, the population of objects in a window is estimated. Changing the window size corresponds to a scale change for fractal measure. By combining measures at several different scales, fractal measures for SAR data are estimated.