A MapReduce Based Visual Workflow Remote Sensing Parallel Product Platform And Related Key Technology

After sent to satellite ground receive station, remote sensing images need to be processed beforeapplied to various industries. However remote sensing images mean a huge amount of data. It is a remotesensing applications "bottleneck" problem to make mass of remote sensing information products produceautomatically in batches. At this point, the paper designs a visualized remote sensing parallel processplatform based on parallel process and MapReduce technology and aim to realize a more common mode toresolve remote sensing parallel processing problems. The platform is planned to be used to process massremote sensing data.In this article, the main contribution and innovations are listed following:Firstly, it is a workflow integrated platform. An idea of taking advantage of a workflow to refinealgorithms is proposed. The idea tends to call an algorithm as a function. Also it tends to customizealgorithms by a visualized method. As a guide, the algorithm workflow is designed and a visual editingalgorithm organizational environment is realized, aiming to save repeated developed time. It makes easierand flexible to organize an algorithm and configure it.Second, it is a Hadoop-power integrated platform. A combination of Hadoop is applied which is such ahot spot. With Hadoop, we get a highly promotion in data storage capacity and mass data processingcapacity. There is also a high improvement in parallel processing capacity、expansibility and data safety.Thirdly, a MapReduce based remote sensing algorithm developing model is proposed. A data interfacefor remote sensing data is designed based on MapReduce framework. The interface is to turn a paralleldevelopment which may be complicated to a much easier develop process. Through the interface, thealgorithm developer may focus on algorithm model itself without taking care of the requirement ofMapReduce for the data structure.Finally, it is based on experiments and verification. Through a large number of experiments, theplatform in the paper is proved to get improved in parallel processing and other performances in a pluralityof perspectives.