A Remote Sensing Products Processing System Based On Hadoop

Satellite remote sensing is an important method for earth observation and research of global environment changes. Status information of forest, grassland, water etc. can be acquired through the use of ecological environment remote sensing products. These products have been widely used in scientific research institutes and industry sector. Meanwhile, lots of countries have launched a plenty of remote sensing satellite, such as Landsat, Terra, Aqua, Spot etc., which producing huge amounts of earth observation data every day. However, the production of global ecological remote sensing products has been considered as a compute-intensive and data-intensive task because of the massive input data and complex calculation model. Traditional serial production method is unable to meet the requirements of the task since its low computation efficiency.To solve this problem, a remote sensing products processing system is presented to effectively produce the ecological environment remote sensing products by utilizing Hadoop, which is a quite popular distributed computing framework. The main contents of this thesis is list as below.(1) This thesis implementate the algorithms of the Global Environment Monitoring Index and Global Grassland Drought Index by using MapReduce programming model. The format of input key, input value, and partition function are elaborately designed by taking the operational principle of MapReduce model and the characteristics of the algorithms into account. To deal with the complicated production work which contains several MapReduce procedures, workflow technology is used in order to tackle the dependence between these procedures.(2) Different computation methods, including the serial and distributed computation methods, are used to processing the global scale of the products mentioned above. This thesis compared the efficiency differences between serial and distributed computation, and also measured the efficiency of Hadoop cluster by using different number of computing nodes in the process of the production.(3) This thesis design and implement the Global Ecological Environment Remote Sensing Production System by utilizing the technologies of Hadoop, J2 EE, WebGIS, etc. The system carries out the production tasks in Hadoop cluster according to the requirement of users. The functions of order analysis, production control, data and product management are integrated into the system and form a relatively complete production environment. Compared with the traditional desktop remote sensing production system, the presented system improves the efficiency of production, and also provides a convenient and efficient way for users interested in the ecological environment to acquire the related products.