葉從勝，胡曉兵， 管清貴， 殷國富（四川大學制造與工程學院，四川成都 610065） 摘 要：針對傳統(tǒng)天然氣檢測系統(tǒng)存在的檢測效率低和準確性不高的問題，提出了基于顯著形狀特征的天然氣表數(shù)字圖像識別方法，通過對數(shù)字外輪廓進行簡潔、高效的編碼，并輔助數(shù)字長寬比、內腔等其他形狀特征，從而能同時識別整字和半字出現(xiàn)的情況，即而實現(xiàn)了氣表檢測過程中讀數(shù)的自動化。實踐證明，新系統(tǒng)的檢測效率大為提高。 關鍵詞：氣表檢測；顯著形狀特征；數(shù)字識別 中圖分類號：TP391.4 文獻標識碼：A 文章編號：1672-4984（2007）01-0051-03 Natural gas meter inspection based on image processing YE Cong-sheng， HU Xiao-bing， GUAN Qing-gui， YIN Guo-fu （School of Manufacturing and Engineering，Sichuan University，Chengdu 610065, China） Abstract:This paper introduced a approach to identify numeric reading of natural gas meter，which was based on the representative shape characters of printed digits.By encoding the outer contour of digit with a compact and highly effective method which was aided by other shape characters，such as the ratio of length-width and the lumens of some digits，both results could be recognized correctly when either the half-digit or whole-digit situation appeared.So the automation of reading the natural gas meter in the process of inspection was implemented.After applying the research introduced in this paper to practice，the new system has greatly improved the efficiency，from the half automation system of inspecting 8 natural gas meters per day to 40 now. Key words:Natural gas meter inspection；Representative character；Digit recognition
