Study On The Architecture Of Satellite Remote Sensing Images Preconditioning System And Preconditioning Algorithm

With keeping development of space technology, processing information on satellite is coming into practice. To achieve the goal, based on the analysis of processing remote sensing data from satellites, the parallel information processing architecture suitable for satellite environment should be researched and designed, with correlative parallel algorithms.Based on the analysis of the remote sensing images preprocessing, a parallel architecture of remote sensing preprocessing system suitable for satellite environment is presented. And under the framework of entire architecture, with the parallel operation progress, the composition of parallel processing node, the storage organization of geometric correction array and the topological structure of system is analyzed and designed in details.In the architecture, the processing unit dedicated for satellites adopts Chip Multi-Processor Architecture (CMP), which has the process to be paralleled within the processing nodes, and so result in the decrease of system scale with high processing performance kept. In this paper, based on the Cell processor and its Cell Broadband Engine Architecute (CBEA), the design of CMP for satellite is discussed in details, especially on Synergistic Processor Unit.As the processor dedicated for satellites adopts CBEA, to efficiently imbalance the load of each parallel processing unit, the Image Cutting Algorithm base on Checking Intersection-point is introduced and implemented. With the result data, the performance of algorithm is analyzed.Base on the formal design of system architecture, processor architecture, And the image cutting algorithm, CMP-Parallel Geometric Correction Algorithm is presented, which suits CMP architecture, and the detail progresses of algorithm on the CMP is introduced. The algorithm is carried out on the CBEA simulator, and with the result data, the algorithm efficiency is analyzed, while efficiency influences of optimizing methods for CMP are discuessed.