Abstract:To meet the demand for rapid map tile generation from GB/TB-level remote sensing imagery, this study proposes an efficient tile generation method based on Celery, tailored for domestically developed software and hardware platforms. An adaptive task partitioning strategy is first introduced to accommodate irregular remote sensing images. Tiling tasks are then divided into basic and composite task groups according to the tile levels. Basic tasks directly slice the original imagery based on its spatial extent, while composite tasks generate high-level tiles by synthesizing low-er-level ones. To mitigate memory overflow at computation nodes, a memory optimization strategy based on quantitative consumption is employed. Experimental results show that the proposed method achieves up to a 6.45 times improvement in tiling speed over GeoServer in a 7 nodes environment, demonstrating strong practicality and efficiency.