Research And Development Of Fast Pyramid Generation Algorithm For Massive Remote Sensing Image

With the rapid development of remote sensing technology, applications using remote sensing data cover more areas, like terrain mapping, digital earth, weather forecast and resource survey. High-resolution, multi-temporal and multi-spectral remote sensing images become the basic data for Earth Observation (EO).And the volume of the remote sensing data grows exponentially.The method to retrieve and display the user interested data efficiently in the massive remote sensing data is the research hot spot. The image tiling, Level of Detail (LOD) and image pyramid model were introduced. Combining both image tiling and LOD’s advantages, image pyramid model becomes the routine way to organize remote sensing data. Image pyramid is generated by first smoothing the image with appropriate smoothing filter and then sub-sampling the smoothed image, to produce different resolutions and scales images. This method has low efficiency when applied to massive remote sensing images and does not meet the real-time updating requirement for web map service. The parallel algorithm improves the computation ability for massive datasets, especially for the data intensive task—pyramid generation.The parallel pyramid generation algorithm is produced in the paper taking advantage of distributed data storage and parallel computation technology. The original remote sensing data is divided N*M grid based on tiles. The traditional big construction job is divided into N*M small jobs. The job scheduler is developed. The linear quadtree index oriented pyramid is also implemented to organize all pyramid tiles scattered in the cluster together. This index can solve the problems like ambiguousness in latitude/longitude. The algorithm is implemented under the cluster with mater node and five slave nodes. The results of our experiment show that the algorithm introduced in the paper is more scalable than traditional algorithm.