Research On Remote Sensing Image Parallel Computing Strategy

With the increase of the channels in remote sensing information access and the develop of sensor technology, the data volume and complexity of remote sensing image also grow rapidly, and high-speed, high precision become new space application requirements of remote sensing data processing. It becomes natural and appropriate to design a general model of parallel computation which is appropriate for mass remote sensing data processing.According to the level division of digital image processing, taking the satellite remote sensing image processing for example, paper summarizes the general flow of remote sensing image processing, and divides it into five levels, including preprocessing level, data processing level, information extraction level, the knowledge level and application level. According to the relation among pixels in remote sensing image processing algorithm, it can be divided into three categories, including point-processing algorithms, local processing algorithm and global processing algorithm. Based on the basic idea of parallel programming, paper designs the general purpose parallel computing models of the remote sensing image processing aiming at three algorithms, and taking image fusion, filtering and segmentation for example, experiments and analyses performance of the general parallel computing model. Focusing on remote sensing image segmentation algorithm in global processing algorithms in-depth study, propose a kind of general parallel computing strategy applicable for global algorithm. Experiments show that the proposal of remote sensing image parallel computing general model, which can effectively improve the speed and efficiency of remote sensing image processing, makes real-time requirements of remote sensing applications greatly satisfied.