基于CUDA加速的X射线安检机图像处理算法研究  

Image Processing Algorithms on X-ray Security Machine Based on CUDA Acceleration

作　　者：黄旭南 项安[1] 

机构地区：[1]同济大学电子与信息工程学院,上海201804

出　　处：《机电一体化》2015年第8期3-8 46,共7页Mechatronics

摘　　要：针对X射线安检机处理大分辨率图像时出现的卡顿问题,进行了在显卡上实现安检图像处理算法的GPU并行化研究。利用NVIDIA公司的显卡进行GPU并行计算,通过分析原本工作在CPU上的串行算法,结合NVIDIA提供的CUDA编程工具,使用大量的CUDA线程并行处理图像的每个像素点,对安检机图像处理算法进行了并行化加速,充分发挥了GPU的高浮点运算能力和并行特性,使得GPU算法效率大大超越CPU算法。在准确计算出安检机所需的图像的前提下,计算速度比原CPU算法提升了近50倍。This paper raised the X-ray image security machine processing algorithms based on GPU parallel acceleration,which could reduce the computing time when the machine dealing with large-resolution image.With NVIDIA's GPU graphics card,by analyzing the original serial algorithm work on the CPU,combined with NVIDIA CUDA programming tools,we use a large number of CUDA threads to process each pixel image,and implement the GPU parallel algorithm.Because of the high floating point capability and the parallel nature of the GPU,the GPU algorithms is faster than the CPU algorithm to enhance the nearly 50-fold,which compute the same image and work out with the same result.

关 键 词：GPU 并行计算 X射线 安检机 图像处理算法 CUDA 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]