欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
针对高光谱几何校正计算复杂,大数据量频繁传输降低处理效率,无法满足实时需求等问题,提出基于多核CPU和GPU的并行计算模型。实现基于GPU的并行几何校正,并引入流水线并行思想提出基于多线程的数据读写优化方法,实现重采样部分的数据I/O优化。应用航空推扫成像仪所得高光谱数据进行实验,验证该方法能够有效地隐藏部分硬盘与内存间的数据I/O时间,几何校正加速比达到4.03,在基于GPU的并行计算基础上提高了1.74倍。
英文摘要：
Aiming at the quick processing of geometry correction,a parallel computation model based on multicore CPU and GPU is proposed in consideration of the complex computation and frequent IO operation during geometry correction.GPU based parallel geometry correction was implemented.By introducing the technique of stream parallel,an optimized method of data IO is proposed to realize the data IO optimization of re-sampling section.The experiment of airborne hyperspectral data collected by push-broom hyperspectral imager was executed.The result shows that the geometry correction processing after optimization can effectively hide the data IO time of partial hardware and internal memory.The speedup ratio of geometry correction reaches 4.03,incresed by 1.74 times of GPU based parallel computation.
同期刊论文项目
同项目期刊论文
Development of a dual-path system for band-to-band registration of an acousto-optic tunable filter-b
Combined laser and broadband light for measuring the tuning relation of an acousto-optic tunable fil
Simulation of hyperspectral radiance images with quantification of adjacency effects over rugged sce
期刊信息
位置：