泥沙研究2025，Vol.50Issue(6)：16-22,7.DOI:10.16239/j.cnki.0468-155x.2025.06.003
基于深度学习的细颗粒泥沙絮团测量及处理系统研究
Research on the measurement and processing system of fine sediment flocs based on deep learning method
摘要
Abstract
To address the limitations of traditional image processing methods in measuring fine sediment flocs,which are susceptible to interference from out-of-focus particles,low efficiency,and significant subjectivity in parameter selection,this study proposes an automated image processing model(YOLO-floc)based on the YOLOv7 deep learning framework.A high-resolution CCD camera system was integrated with the YOLO-floc model to achieve precise floc boundary localization and effective elimination of defocused particles,thereby ob-taining critical parameters including floc size and fractal dimension.Two validation experiments demonstrated that the model exhibited a high degree of consistency with manual measurements in different size distribution of glass beads and microplastic particles.Furthermore,the model successfully quantified the evolutionary proces-ses of floc size during montmorillonite particle flocculation.The experimental results validate the reliability of this methodology under laboratory conditions.The application of deep learning models to sediment floc analysis provides an efficient and high-precision technical solution for studies on sediment transport and pollutant migra-tion mechanisms.关键词
絮团测量/YOLOv7模型/深度学习/目标检测/图像处理Key words
floc measurement/YOLOv7 model/deep learning/object detection/image processing分类
建筑与水利引用本文复制引用
叶宏辉,许春阳,陈永平,谢嘉钰..基于深度学习的细颗粒泥沙絮团测量及处理系统研究[J].泥沙研究,2025,50(6):16-22,7.基金项目
国家重点研发计划项目(2023YFC3008100) （2023YFC3008100）
国家自然科学基金项目(42376171) （42376171）
中央高校基本科研业务费专项(B230201047) （B230201047）
国际(地区)合作与交流项目(W2421060) （地区）
