基于均匀驱动计算架构的卫星影像快速Mask匀光方法研究  

Research of Fast Mask Dodging for Satellite Image Based on CUDA

作　　者：常高[1,2] 王华斌[2] 李国元[2] 刘金刚[1,3] 

机构地区：[1]首都师范大学计算机科学联合研究院,北京100048 [2]国家测绘地理信息局卫星测绘应用中心,北京101300 [3]中国科学院计算技术研究所,北京100190

出　　处：《科学技术与工程》2014年第12期240-244,共5页Science Technology and Engineering

基　　金：科技部科技支撑项目(2011BAB01B05);国家高技术产业化项目资助

摘　　要：针对大数据量的卫星遥感影像进行匀光处理的过程计算量大,效率低的情况,提出了一种基于均匀驱动计算架构(compate unified device architecture,CUDA)平台的卫星影像快速匀光并行处理方法。根据现有算法结构上的特点,将处理过程中的主要运算部分交由GPU完成。实验通过对资源三号卫星所得影像做匀光处理,对比了GPU与多核CPU运行速度和效果,分析了该方案的加速性能和实用性。实验表明,在保持原处理质量的同时,该方法构建的并行处理模型有效地提高了原有匀光算法的处理速度。Due to large size and high definition, the process of removing uneven illumination of satellite remote sensing image is higher time-consuming and lower efficiency than before. Based on CUDA platform, a parallel method of removing uneven illumination of satellite image is presented. According to the feature of the given algorithm, GPU performed as the main computing unit. The experiment compares the processing speed with that of multi core CPU, analyses the performance of speeding up. The result demonstrates that with this parallel model built, by remaining processing quality, the original algorithm of removing uneven illumination has been speeded up effectively.

关 键 词：遥感影像 Mask匀光法 GPU CUDA 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]