协同CPU和GPU的核密度估计及其可视化算法  

Kernel Density Estimation and Its Visualization Algorithm Combining CPU and GPU

作　　者：胡森 高苏 蔡忠亮[1] HU Sen;GAO Su;CAI Zhongliang(School of Resource and Environmental Sciences,Wuhan University,Wuhan 430079,China;Yunnan Provincial Mapping Institute,Kunming 650034,China)

机构地区：[1]武汉大学资源与环境科学学院,湖北武汉430079 [2]云南省地图院,云南昆明650034

出　　处：《地理空间信息》2024年第6期29-33,47,共6页Geospatial Information

基　　金：国家重点研发计划资助项目(2021YFB2501101)。

摘　　要：大数据时代背景下,空间数据点规模越来越大,图像分辨率越来越高,使用CPU计算核密度估计结果并对其可视化的效率越来越低,难以满足应用对实时性的需求。针对该问题,提出了一种协同CPU和GPU的核密度估计及其可视化算法,该算法结合CPU的控制能力、GPU的并行计算能力以及OpenGL中的核心模式,并借助显存映射,同时优化了核密度估计的计算和可视化2方面。实验结果表明,相较于CPU并行和串行算法,该算法的执行效率分别提高了约5倍和20倍,且随着图像分辨率的提高,加速比呈现逐步上升的趋势。In the context of the big data era,the scale of spatial data points is getting larger and larger,and the image resolution is getting higher and higher,thus,the efficiency of using CPU to calculate kernel density estimation result and visualize it is getting lower and lower.It’s difficult to meet the real-time requirements of applications.Aiming at this problem,we proposed a kernel density estimation and visualization algorithm combining CPU and GPU.By utilizing the control ability of CPU,the parallel computing ability of GPU and the core mode in OpenGL,this algo-rithm optimizes both computational and visual aspects of kernel density estimation.The experimental results show that compared with CPU based parallel and serial algorithms,the efficiency of this algorithm is increased by about 5 times and 20 times respectively,and with the improve-ment of image resolution,the acceleration ratio shows a gradual upward trend.

关 键 词：核密度估计 可视化 GPU OPENGL 统一计算架构 

分 类 号：P208[天文地球—地图制图学与地理信息工程]