Design And Realization Of Remote Sensing Image Fusion Distributed Parallel Computing System Based On Globus Toolkit

With the development of remote sensing technology and sensor technology, multi-platform, multi-sensor, multi-temporal, multi-spectral, multi-angle and multi-resolution remote sensing image data is quickly coming in an alarming number. However, due to the different imaging principle and application purpose of a variety of remote sensing images.There are some limitations and differences in the geometric, spectral resolution and spatial resolution of any remote sensing data from a single sensor, which is difficult to meet the actual needs. To solve this problem, thus, multi-sensor image data fusion technology emerges. After image fusion, a new image is generated by fusion multiple images from a variety of sensors imaging the same scene. Image fusion effectively improves the credibility and intelligibility of the image, while reduce the ambiguity of the image, which is more conducive to human vision or the computer to detect, classify, identify, understand and other processing. Image fusion technology includes three levels:pixel level, feature level and decision level, in the paper, pixel-level image fusion algorithm is studied. Firstly, the theoretical basis of traditional fusion algorithms is analyzed, and the strengths and limitations of each algorithm are compared combined with experimental results. On this basis, we focus on the image fusion method based on IHS transform and wavelet transform, and present a modified fusion algorithm which is a combination of wavelet transform and IHS transform. Secondly, the quality evaluation methods of fused image are analyzed, and evaluated the quality of experimental results of various algorithms implemented in the paper. Experimental results show that the method presented in the paper can effectively reduce the spectral distortion of the fusion results. It greatly improves the spatial resolution, and at the same time it is better to retain the spectral information of multi-spectral images.The convenient and rapid access to mass remote sensing data make the image processing increasingly important, meanwhile the accuracy and speed of image processing is demanded more and more. These bring a new challenge to image processing technology. Remote sensing image processing is a data-intensive, computing-intensive operation, and the traditional stand-alone mode has been difficult to meet the capacities of data storage, transmission, processing and others required by image processing. The emergence of distributed parallel processing system brings hope to solve the problem. This computing system with its powerful computing capacity, unique flexibility and low cost become the effective protection of mass data storage, transmission, analysis and management. The paper analyzes the problems of traditional distributed parallel system and the application status of the system in the field of remote sensing data processing, and proposes that with the distributed parallel processing platform of Globus grid, distributed parallel system which is suitable for remote sensing image data processing is constructed. We use the grid middleware to realize the functions such as resource dynamic monitor, node isomerism shield, optimizing the resource selection and working in coordination with calculations. Based on the architecture of Globus Toolkit, the paper installs and configures Globus grid development environment, design the suitable parallel mode, and achieve the function modules of task distribution and receiving of the system. Finally, we compare the experimental efficiency in stand-alone mode and distributed parallel system mode. The experimental results show that distributed parallel system greatly reduces the processing time and improves processing efficiency. Thus, grid-based distributed parallel system is a better choice for mass data storage, transmission and fast processing and other issues.