The Design And Implementation Of Disaster Information Extraction For Parallel Algorithm Of SAR Image

With the development of remote sensing technology, application of remote sensing is increasingly common in various fields, our current sensing applications are in the the stage of development,it is necessary to monitor dynamicly long-term for the survival of mankind’s natural environment,extracting disaster information effectively. Synthetic Aperture Radar(Synthetic Aperture Radar- SAR) is a high resolution imaging radar,with the day and all-weather features, has a unique advantage in detecting disaster, but because of SAR images generally contain huge amounts of data information, and SAR image disasters algorithm model are complexity, so SAR image processing algorithm will spend a lot of time for disaster. Usually must be based on real-time images quickly calculate the results obtained in the treatment so as to formulate a workable disaster protection measures, so a quick treatment is necessary to study and design disaster detection algorithm of SAR images, and integrated it intoSAR image processing systems.This paper studies the disaster information extraction algorithms for parallel resistance, and were given a parallel processing scheme based on two different forms of MPI + OpenMp and OpenCL, and integrated the designed algorithm module into SAR image processing software designed, the main work is as follows:(1)Designing a parallel hybrid MPI plus OpenMp shared storage cluster based algorithm. Algorithm design for the road damage multiple nodes and internal levels parallel program,. MPI+OpenMp damaged roads extraction algorithm with hybrid parallel implementations of SAR image data for a detailed data block, and provides the best coarse-grained parallel programs for node part of the internal parallel algorithms.Finally, the experimental analysis of the impact of the number of nodes on the hybrid model algorithm proved computer cluster of hybrid parallel with OpenMp program in a small test chamber which is relative to the superiority of the serial algorithm.(2)Taking into account the GPU advantage in SAR image processing on huge amounts of data, this paper designed a parallel algorithm based OpenCL of road damage.In particular the design process, the first part of the algorithm can be optimized forperformance in parallel. Optimized algorithm based on the detailed design of the storage and thread partitioning scheme, obtained by experimental tests up to speedup of 13 times.(3)Developing software on image information extraction of the SAR for disaster,object-oriented approach to building the various functional modules in detail design.The separation of algorithm module and the system interface is order to achieve a low coupling of the entire system. For each disaster processing algorithms for different hardware environment with OpenCL or OpenMp + MPI parallel acceleration, and made the algorithm integrate into the software system with users used friendly.