基于图像识别的床面推移质分布密度检测方法  

The Method for Detecting Bed Load Distribution Density Based on Image Recognition

作　　者：高昂[1] 刘明潇[2] 孙东坡[1] 曹震[1] 陈云飞[1] 

机构地区：[1]华北水利水电大学水利学院,河南郑州450045 [2]西安理工大学水利水电学院,陕西西安710048

出　　处：《人民黄河》2015年第9期20-23,共4页Yellow River

基　　金：国家自然科学基金重点项目(51039004);华北水利水电大学大学生创新计划资助项目

摘　　要：为研究非均匀推移质的输移特性,开发了由影像采集系统和图像识别软件处理系统组成的床面运动颗粒分布密度及组成检测系统。该系统采用非接触式影像采集方式,运用图像识别技术提取颗粒特征值,对水沙运动零干扰。图像识别软件采用中值滤波及卷积处理等方法进行图像增强,采用自适应阈值技术分割图像。推移质低强度输移时,图像识别精度较高;推移质输移强度较高时,识别精度可通过修正系数进行修正。同时,还可以通过图像识别获取床面推移质的级配组成。水槽试验与检测对比结果表明:该检测系统具有较高的识别可靠性。For studying non - uniform transport properties, a system consisted of image acquisition system and image recognition software processing system for detecting distribution density and composition of bed surface particles was presented. The detection system adopted non-contact image acquisition mode and utilized image recognition technology to extract particle characteristic value and had no interference with water and sediment movement. Software system used median filter and convolution processing for image enhancement, as well as used the adaptive threshold to segment grayscale. When bed load transport intensity was low, image recognition precision was high ; when bed load transport intensity was high, the identification accuracy could be corrected by introducing correction coefficient. Meanwhile, this system could also ascertain the size distribution composition of bed load based on image recognition technology. The flume experiment and the detection comparison results show that the detection system has higher recognition reliability.

关 键 词：推移质 分布密度 泥沙级配 图像识别 

分 类 号：TV149[水利工程—水力学及河流动力学]