作者:
孫廣旗(新興河北工程技術有限公司,河北 邯鄲 056107)
崔 勇(中國科學院沈陽自動化研究所,遼寧 沈陽 100169)
申振鑫(新興河北工程技術有限公司,河北 邯鄲 056107)
王 宇(中國科學院沈陽自動化研究所,遼寧 沈陽 100169)
摘要:本文針對鑄管行業常見的鑄管鑄字號的特點,設計了基于深度學習的鑄管內壁陽文鑄字的檢測和識別方案。文中對比了主流的神經網絡結構,分析其優缺點并確定了方案的細節。在鑄管廠部署實施后,通過科學的方法對應用效果進行了統計,驗證了該鑄字檢測和識別方案的優異效果。
關鍵詞:CNN;深度學習;鑄字檢測;鑄字識別
Abstract: In this paper, according to the characteristics of common ironpipecharacters, we design the detection and identification scheme ofraised characters in the inner wall of iron pipes based on deep learning.In this paper, we compare the most famous neural network structures,analyze their advantages and disadvantages and determine the details ofthe scheme according the result. After being deployed and implementedin the ductile iron pipes factory, the application effect is statisticallyanalyzed scientifically, and the excellent effect of the casting detectionand identification scheme are verified.
Key words: Deep learning; CNN; Iron-Pipe character detection; Iron-Pipe character Recognition
在線預覽:基于深度學習的鑄管鑄字的檢測與識別方案設計.pdf
摘自《自動化博覽》2021年10月刊