基于无人机遥感平台的砾石自动筛分方法  

Automatic gravel sieving method based on UAV remote sensing platform

作　　者：高仕赵[1] 毛海英 范庆来[1] 冀自青[3] GAO Shizhao;MAO Haiying;FAN Qinglai;JI Ziqing(School of Hydraulic and Civil Engineering,Ludong University,Yantai 264025,Shandong China;School of Civil andEngineering and Architecture,Guangxi University of Science and Technology,Liuzhou 545006,Guangxi China;State KeyLaboratory of Hydraulic Engineering Intelligent Construction and Operation,Tianjin University,Tianjin 300072,China)

机构地区：[1]鲁东大学水利土木学院,山东烟台264025 [2]广西科技大学土木建筑工程学院,广西柳州545006 [3]天津大学水利工程智能建设与运维全国重点实验室,天津300072

出　　处：《华中科技大学学报(自然科学版)》2025年第3期23-30,共8页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：山东省中青年专家奖励基金资助项目(BS2014SF016);山东省自然科学基金资助项目(ZR2019MEE010)。

摘　　要：针对砾石滩涂颗粒的粒度分布测量问题,提出了一种基于无人机遥感平台的砾石自动筛分方法(GAS).首先,利用无人机悬挂已知尺寸的标定球体进行远距离图像采集;然后,根据砾石图像的几何形状的凸特征,确定了砾石像素斑块的最大搜索范围;其次,依据砾石在颜色域空间中的特点对砾石像素斑块进行搜索和判断;接下来,依据像素斑块搜索的结束方式来确定砾石图像生成方式;最后,应用ImageJ软件手动测量方法对GAS方法进行验证.结果显示:GAS方法所得结果与手动测量结果的相关系数为96.3%~99.6%,相对误差为4.5%~29.7%,说明GAS方法在粒径测量方面具有较高的准确性和可靠性.此外,还与自动图像分割法(AIF)和BASEGRAIN方法进行了对比,结果显示:GAS方法生成的级配曲线相对误差为4.5%~29.7%,而AIF和BASEGRAIN方法生成的级配曲线的相对误差分别为10.9%~84.6%和23.1%~99.9%,说明GAS方法在实际图像数字筛分中具有更高的分割精度和更好的稳定性.Aiming at the problem of measuring the particle size distribution of gravel beach particles,an automatic gravel sieving method(GAS) based on an unmanned aerial remote sensing platform was presented.First,a calibrated sphere of known size was suspended by a unmanned aerial vehicle(UAV) for remote image acquisition.Second,the maximum search range of gravel pixel patches was determined based on the convex features of the gravel image geometry.Third,the gravel pixel patches were searched and judged based on the characteristics of gravel in the color space.Fourth,the generation method of gravel images was determined based on the end method of pixel patch search.Fifth,the grain size was measured manually by applying ImageJ.The results show that the correlation coefficients between the results obtained by the GAS method and the manual measurements are 96.3%~99.6%,and the relative errors are 4.5%~29.7%.This result indicates that the GAS method has high accuracy and reliability in particle size measurement.In addition,the results were compared with the automatic image filtering method(AIF) and the BASEGRAIN method.The results show that the relative errors of the cumulative curves generated by the GAS method are 4.5%~29.7%.In comparison,the relative errors of the cumulative curves generated by the AIF and BASEGRAIN methods are 10.9%~84.6% and 23.1%~99.9%,respectively.This result indicates that the GAS method has higher segmentation accuracy and better stability in realimage digital sieving.

关 键 词：无人机 砾石 自动筛分 级配曲线 图像处理 粒径 

分 类 号：TV21[水利工程—水文学及水资源]