基于GPU并行算法的景观斑块周长模拟测量仿真  

Simulation of Landscape Patch Perimeter Measurement Based on GPU Parallel Algorithm

作　　者：吴国荣[1] 渠琛 WU Guo-rong;QU Chen(Nanchang University,Nanchang Jiangxi 330000,China)

机构地区：[1]南昌大学,江西南昌330000

出　　处：《计算机仿真》2023年第3期280-284,共5页Computer Simulation

摘　　要：针对景观斑块周长模拟测量过程中收敛速度较慢的问题,提出基于GPU并行算法的景观斑块周长模拟测量方法。通过航拍技术采集景观斑块航拍图像,利用纠偏处理过程与高斯滤波处理过程对景观斑块航拍图像进行预处理。将景观斑块航拍图像输入GPU内,通过计算机GPU片段程序构建景观斑块航拍图像的高斯差分金字塔模型,并行运算过程提取景观斑块航拍图像特征点,确定景观斑块航拍图像中景观斑块的边界线。以边界线为基础,计算景观斑块周长。仿真测试结果显示,上述方法特征点提取过程中横轴与纵轴误差完全收敛时间分别为2.12s和1.80s,周长测量绝对误差与相对误差分别控制在0.60m和1.00%以下。In order to improve the convergence speed of landscape patch perimeter simulation measurement,a landscape patch perimeter simulation measurement method based on GPU parallel algorithm was designed in the paper.Aerial images of landscape patches were collected through aerial photography technology.The correction process and Gaussian filtering process were applied to preprocess the aerial image of landscape patch.Aerial images of landscape patches were input into GPU.Based on the computer GPU fragment program,the Gaussian difference pyramid model of aerial image of landscape patch was established.The operation process was parallel to extract the feature points of landscape patch aerial image,determining the boundary line of landscape patch in landscape patch aerial image.According to the boundary line,the perimeter of the landscape patch was obtained.The simulation results show that the horizontal and vertical axis errors of this method are small(the complete convergence time is 2.12s and 1.80s respectively),and the absolute and relative errors of perimeter measurement are less than 0.60m and 1.00%,respectively.

关 键 词：并行算法 景观斑块周长 模拟测量 航拍图像 特征提取 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]