基于图像识别的土石坝坝料颗粒级配检测方法  

Image recognition based particle gradation detection method for earth and rockfill dam materials

作　　者：罗枭雄 金伟 赵鹏远 钟志鸿 李海波 胡宇翔[1,4] LUO Xiaoxiong;JIN Wei;ZHAO Pengyuan;ZHONG Zhihong;LI Haibo;HU Yuxiang(College of Water Resources and Hydropower,Sichuan University,Chengdu 610065,Sichuan,China;PowerChina Chengdu Engineering Corporation Limited,Chengdu 610031,Sichuan,China;Yalong River Hydropower Development Company,Ltd.,Chengdu 610000,Sichuan,China;State Key Laboratory of Hydraulics and Mountain River Engineering,Sichuan University,Chengdu 610065,Sichuan,China)

机构地区：[1]四川大学水利水电学院,四川成都610065 [2]中国电建集团成都勘测设计研究院有限公司,四川成都610031 [3]雅砻江流域水电开发有限公司,四川成都610000 [4]四川大学山区河流保护与治理全国重点实验室,四川成都610065

出　　处：《水利水电技术(中英文)》2024年第12期217-227,共11页Water Resources and Hydropower Engineering

基　　金：国家自然科学基金(42102316);水利部水库大坝安全重点实验室开放研究基金(YK323002)。

摘　　要：【目的】土石坝坝料级配直接影响砾质土压实性、渗透性、压缩性及应力应变关系,是保证坝体稳定安全的重要标准。目前坝料级配的测量主要采用筛分法,但该方法施工效率低下、误差较大,具有很大的局限。为了实现坝料级配参数的智能检测,【方法】利用DIP技术和人工智能算法及大数据理论,建立了一套坝料颗粒级配识别系统。以两河口水电站坝料为依托,采取了规范化的取样标准,得到了具有代表性的坝料样本数字图片,采用了对比度增强和小波变换为一体的手段为坝料的快速准确分割提供了技术支撑,然后采用阈值分割算法进行坝料图像分割,获得了基于图像识别技术的坝料级配曲线,并通过基于BP神经网络的误差修正模型得到了修正后的坝料级配特征曲线。【结果】将使用该方法得到的级配曲线与实测的级配曲线进行比较,发现两者有较高的吻合度,通过该方法得到的级配曲线可以准确检测坝料的各项指标。【结论】建立的坝料颗粒级配识别系统能快速、准确地得到颗粒的级配曲线,克服了筛分法实用性较低的问题,实现了土石坝复杂坝料的快速检测。[Objective]The grading of earth and rock dam materials directly affects the compaction,permeability,compressibility,and stress-strain relationship of gravelly soil,which is an important criterion to ensure the stability and safety of the dam body.At present,the measurement of dam material gradation mainly adopts the sieving method,but this method has great limitations in terms of low construction efficiency and large error.In order to realize the intelligent detection of dam material gradation parameters,[Methods]a dam material particle gradation identification system is established by using DIP technology and artificial intelligence algorithm and big data theory.Relying on the dam materials of Lianghekou Hydropower Station,a standardized sampling criterion was adopted to obtain a representative digital picture of the dam material samples,and contrast enhancement and wavelet transform as a whole were used to provide technical support for the fast and accurate segmentation of the dam materials,and then the threshold segmentation algorithm was used for the image segmentation of the dam materials to obtain the gradation curves of the dam materials based on the image recognition technology,and the corrected gradation curves were obtained by the BP neural network-based the corrected dam material gradation characteristic curve was obtained by the BP neural network based error correction model.[Results]The gradation curve obtained by this method is compared with the measured gradation curve,and it is found that both of them have a high degree of agreement,and the gradation curve obtained by this method can be used to detect the indexes of the dam material arbitrarily.[Conclusion]The established dam material particle gradation identifi-cation system can quickly and accurately obtain the gradation curve of particles,overcoming the problem of low practicality of the sieving method,and realizing the rapid detection of complex dam materials of earth and rock dams.

关 键 词：土石坝料级配检测 数字图像处理 级配修正模型 BP神经网络 水利工程 

分 类 号：TV41[水利工程—水工结构工程]