水利信息化Issue(5)：29-34,6.DOI:10.19364/j.1674-9405.2025.05.005
基于降水自记纸图片的降水数据自动识别提取方法
Automated precipitation data recognition and extraction approach for precipitation autographic record paper images
摘要
Abstract
To automatically identify and extract minute-level precipitation data from rainfall recording charts and fully exploit the value of historical hydrological paper records,this study proposed algorithms based on machine vision and pattern recognition technologies:a precipitation trace extraction algorithm based on color and spatial features,a precipitation trace restoration algorithm based on trace variation characteristics,and a precipitation data generation algorithm based on pixel relative distances.Python was used to develop software for automated identification and extraction of precipitation data.Using 52 012 rainfall recording chart images from 26 stations in Shanghai from 1978 to 1994 as a case study,the software was applied to extract precipitation data.The results showed that the average processing time per image was 43.7 s,with 96.7%of the images automatically recognized.Among the identified daily precipitation data,93.1%of stations had errors within 1.0 mm,and 97.3%of stations had errors within 5.0 mm,demonstrating a significant improvement in the efficiency of precipitation data extraction from rainfall recording charts.关键词
降水数据提取/自动识别/降水自记纸图片/迹线识别/迹线修复Key words
precipitation data extraction/automatic recognition/rainfall recording chart images/trace recognition/trace restoration分类
天文与地球科学引用本文复制引用
俞汇,董仕达,沈玥,秦红..基于降水自记纸图片的降水数据自动识别提取方法[J].水利信息化,2025,(5):29-34,6.基金项目
国家重点研发计划项目(2023YFC3008100) （2023YFC3008100）
