朱峥嵘, 黄亚锋, 赵立营. 一种利用GPU加速的轨迹线热力图生成显示方法[J]. 武汉大学学报 ( 信息科学版), 2022, 47(7): 1035-1042. DOI: 10.13203/j.whugis20200001
引用本文: 朱峥嵘, 黄亚锋, 赵立营. 一种利用GPU加速的轨迹线热力图生成显示方法[J]. 武汉大学学报 ( 信息科学版), 2022, 47(7): 1035-1042. DOI: 10.13203/j.whugis20200001
ZHU Zhengrong, HUANG Yafeng, ZHAO Liying. A Method of Generating and Displaying Trajectory Line Heat Map with GPU Acceleration[J]. Geomatics and Information Science of Wuhan University, 2022, 47(7): 1035-1042. DOI: 10.13203/j.whugis20200001
Citation: ZHU Zhengrong, HUANG Yafeng, ZHAO Liying. A Method of Generating and Displaying Trajectory Line Heat Map with GPU Acceleration[J]. Geomatics and Information Science of Wuhan University, 2022, 47(7): 1035-1042. DOI: 10.13203/j.whugis20200001

一种利用GPU加速的轨迹线热力图生成显示方法

A Method of Generating and Displaying Trajectory Line Heat Map with GPU Acceleration

  • 摘要: 为解决大数据量带来的热力图生成效率低的问题,引入基于图形处理器(graphic processing unit,GPU)的并行计算方法,并结合轨迹线模型,提出了一种利用GPU加速的轨迹线热力图生成显示方法。首先,针对轨迹点分布不均、邻域半径设置不合理等条件下产生的热力值不连续、不均等问题,采用轨迹线模型提升了热力图的效果。其次,针对大规模数据计算产生的热力图生成效率低的问题,通过GPU并行计算并配合内核函数参数调优、循环展开、像素缓冲对象显示等策略大幅提升算法计算效率。实验结果表明,所提方法较传统的基于中央处理器(central processing unit, CPU)的方法计算效率提升了5~30倍,且随着图像分辨率和轨迹数据的增加,算法加速比有逐步上升的趋势。

     

    Abstract:
      Objectives  To deal with the high computation expense in data-intensive heat map generation, parallel computing technology based on graphic processing unit(GPU) is introduced to the heat map generation filed. With combination of trajectory line model, we propose a novel method for trajectory line heat map generation and display based on GPU.
      Methods  Firstly, in order to avoid discontinuity and inhomogeneity of heat value resulted from different dense of trajectory points or unreasonable neighborhood threshold, trajectory line model is adopted to improve the effect of generated heat map. Secondly, aiming at the problem of low efficiency of heat map generation caused by large-scale data calculation, the proposed method can improve the efficiency greatly via a combination of GPU parallel computing, tuning of kernel function parameter, loop unrolling and pixel buffer object.
      Results  Experimental results show that processing speed of the proposed method is from 5 to 30 times faster than CPU(central processing unit))-based implementation.
      Conclusions  Additionally, the accelerator ratio tends to be higher as the resolution of heat map or the increase of trajectory data.

     

/

返回文章
返回