水利信息化Issue(5)：60-67,8.DOI:10.19364/j.1674-9405.2025.05.010
基于BM3D算法的大量程水尺水位图像去噪与识别方法研究
Research on denoising and recognition method of large-scale water level images based on BM3D algorithm
摘要
Abstract
In view of the insufficient dynamic adaptability of traditional fixed-focus water level camera monitoring equipment in automatically tracking water level lines of large-range staff gauges,and the problem of false detection in complex environments using image recognition technology,a method for denoising and recognizing large-range staff gauge water level images based on the BM3D algorithm was proposed.A common network dome camera was used as the front-end imaging device,and a stepwise threshold algorithm was employed to capture images of multiple preset points covering the large-range staff gauge.An image denoising model based on the BM3D algorithm was constructed to filter out noise points affecting water level recognition,such as reflections,rain,and snow.Combined with a morphological processing algorithm,redundant connected domain blocks on the images were removed.Gray projection technology was applied to automatically detect and locate the water level line,and image calibration was used to convert pixel distance into actual water level height,realizing automatic tracking of large-range staff gauge water levels.The method was applied at Taoyuan Reservoir in Quanzhou City,Fujian Province,and the water level detection error was within±1 cm,demonstrating high accuracy and reliability in complex environments and providing effective technical support for intelligent water level monitoring.关键词
大量程水尺/水位识别/BM3D算法/图像去噪/连通域区块滤除/灰度投影Key words
large-scale staff gauge/water level recognition/BM3D algorithm/image denoising/connected domain block filtering/gray projection分类
信息技术与安全科学引用本文复制引用
宗泽,傅承琦,吴婷,褚泽帆,蒲升阳..基于BM3D算法的大量程水尺水位图像去噪与识别方法研究[J].水利信息化,2025,(5):60-67,8.基金项目
新疆维吾尔自治区自然科学基金面上项目(2024D01A99) （2024D01A99）
南京水利水文自动化研究所科研立项项目(YJZQ1524002) （YJZQ1524002）
中央级公益性科研院所基本科研业务费专项资金(Y524014) （Y524014）
