基于全局阈值分割的泥沙粒度分布测量及校正  

Measurement and correction of sediment particle size-grading based on global threshold segmentation

作　　者：耿晶晶 唐立模[1] 陈红[1] 林青炜 刘双洪 GENG Jing-jing;TANG Li-mo;CHEN Hong;LIN Qing-wei;LIU Shuang-hong(State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering,Hohai University,Nanjing 210098,China)

机构地区：[1]河海大学水利水电学院,江苏南京210098

出　　处：《泥沙研究》2024年第5期25-33,共9页Journal of Sediment Research

基　　金：国家自然科学基金项目(51479070);浙江省河口海岸重点实验室开放基金项目(ZIHE21008)。

摘　　要：针对泥沙颗粒灰度图像中的噪声和边缘模糊问题,提出了一种新的基于全局阈值分割方法。通过利用背景像素的正态分布特征和像素灰度值的突变点作为参考阈值,实现了有效的阈值分割,进而采用SURF检测和FLANN匹配方法验证了图像分割的准确性和可靠性。构建了泥沙颗粒形状测定系统,通过线性修正方法对系统误差进行了修正,将误差范围控制在±0.01 mm以内。进一步将该方法与马尔文激光粒度仪法进行了比较分析。结果表明:在D50—D60尺度下,两种方法的测量结果基本一致;当颗粒粒径小于D50时,测量结果偏大,大于D60时,测量结果偏小,平均百分位偏差在±0.02 mm范围内,相对百分位偏差在7%以内,证明了该方法的可靠性。This study introduces a novel global threshold segmentation method to overcome noise interference and edge blurriness in grayscale images of sediment particles.By taking the normal distribution characteristics of the background pixels and the abrupt changes in pixel grayscale values as reference thresholds,the effective threshold segmentation was achieved,and the accuracy and reliability of the image segmentation were validated by SURF detection and FLANN matching methods.A sediment particle shape determination system was constructed,and system errors were corrected and controlled within±0.01 mm by using a linear correction method.Through a comparative analysis with the Malvern laser particle size analyzers;the results show that the measurements of both methods are essentially aligned at D50 to D60 scale.For sizes below D50,the measurements by image segmentation method tend to be larger for particle size below D50,while the measurements tend to be smaller for particle size above D60.Overall,the global threshold image segmentation method is reliable,with the average percentile deviation within±0.02 mm and the relative percentile deviation within 7%.

关 键 词：图像识别 全局阈值分割 泥沙粒径级配 激光粒度仪 

分 类 号：TV149[水利工程—水力学及河流动力学]