Parallel Fusion for Remote Sensing Images Based on Grid Services

The bandwidth of network has been greatly improved owing to the development of network,a large number of remote sensing images released on the network as shared resources which could be accessed online.And with the development of network,the traditional remote sensing image processing methods are not well adapted to the new resource-sharing model,so,it is an easy choice to make improvement on software functions.Gird computing is being more attractive environment for parallel image processing on web because of the high bandwidth and plentifully shared computing resources available on the Grid.Grid computing is a resources sharing environment on the web,and resources dynamic changing is an essential characteristic of the Grid,not only the usage of the resources is changing but also new computing resource can join or leave dynamically.As the Grid is a frequently varied system,data updating,nodes joining or leaving,the load of node changing and the performance of the network changing will all affect the performance of the Grid services,so only these dynamic information be collected promptly and the execution plan timely adjustment be made that the high performance process could be guaranteed.The main idea is to utilize the Grid architecture WSRF(Web Service Resource Framework),choose a best plan to do parallel image processing according to the status of services,and provide efficient remote sensing image fusion service.In this paper,a service-oriented architecture for parallel remote sensing image on Grid is presented at first,and different kinds of services of the architecture are introduced briefly.There are five components for the architecture,which can trace and record resources dynamic changing on the Grid and provide an environment for parallel remote sensing image processing.Then,a parallel remote sensing image fusion algorithm on the architecture is presented in detail.The algorithm generate optimal parallel execution plan adaptively for the parallel image fusion according to the resource changes on the Grid.The equations for evaluating the costs of different execution plan are also presented.At last,a testing Grid environment is setup and the partial result image of the image fusion of a TM image and a Spot image is illustrated.