摘要
近年来,随着空间遥感技术的发展,使得遥感影像数据呈几何级数增长,遥感影像的处理面临数据量大、密集度高、计算复杂度高和运算量大等问题。在分析最新GPU(图形处理单元)的并行架构和统一计算设备架构(CUDA)灵活的可编程性的基础上,提出了一种基于CUDA的遥感影像的高效处理方法,以遥感影像处理中常用的快速傅里叶变换、边缘检测和模板匹配3种方法为例,详述了基于CUDA的高效并行处理过程,且对不同大小的实际影像进行了实验。实验结果表明,与CPU计算相比,利用CUDA计算能够获得高达10到40倍的加速比,大大的提升了对遥感影像的处理能力。
As the development of space remote sensing technology in recent years witnessed a geometric growth in the data size of remote sensing images. Consequently, the process of remote sensing images is faced with such challenges as large data size, high intensity, high computational complexity and large computational quantity, and so on. Based on the analysis of the parallel architecture of the latest GPU and the flexible programmability of CUDA (Computer Unified Device Architecture), this paper presents an efficient method for processing remote sensing images on the basis of CUDA. This paper takes FFT, edge detection and template matching, three common methods in remote sensing image processing, as examples, and details the efficient parallel processing procedures of them. The experiments on different images with different data size proved that GPU is 10 to 40 times faster than CPU, which is a dramatic progress in remote sensing image processing.
出处
《地理空间信息》
2011年第6期47-54,4,共8页
Geospatial Information
基金
中央高校基本科研业务费专项资金资助项目(6081001)