CPU和GPU协同处理的光学卫星遥感影像正射校正方法

A CPU-GPU Co-processing Orthographic Rectification Approach for Optical Satellite Imagery

系统地探讨基于CPU和GPU协同处理的光学卫星遥感影像正射校正方法。首先使用"层次性分块"策略设计基于CPU和GPU协同处理的正射校正方法,然后通过配置选择优化和存储层次性访问等手段进一步提高方法执行效率。在Tesla M2050 GPU上对资源三号卫星下视全色影像进行正射校正的试验结果表明,本文方法可大幅提高光学卫星遥感影像正射校正效率,与传统串行正射校正算法相比,加速比最高达到110倍以上,相应的处理时间压缩至5 s以内,可满足对海量数据光学卫星遥感影像进行快速正射校正的要求。

A new CPU-GPU co-processing orthographic rectification approach for optical satellite imagery is put forward. First, ＂hierarchical tiling＂ strategy is applied to form the basic algorithm flow then the algorithm performance is further improved in the respect of configuration optimization and memory hierarchical access. The algorithm is applied to the orthographic rectification of ZY-3 nadir panchromatic imagery and the experimental data is collected. By analyzing the data, it is found that the processing time of this algorithm on Tesla M2050 GPU is less than 5 s and the highest speedup ratio to the traditional serial algorithm is more than 110 times. This result demonstrates that this algorithm significantPy improves the orthographic rectification efficiency and fully satisfies the requirement of fast orthographic rectification for large data optical satellite imagery.