摘要
针对铝型材挤压模设计后图纸重用困难、寻找相似型材面模具效率低等问题,提出基于型材面结构特征的匹配算法,根据新型材面图像匹配数据库中相似的型材面图像,寻找对应的模具结构用于参考。该算法利用YOLOv5算法有效识别型材面结构特征,自动标注结构位置,通过计算对应结构在型材面中的位置、面积与类型得出结构相似度,并采用最小费用最大流算法对结构进行一对一匹配,以获得最佳匹配结果。
In view of the difficulties in drawing reuse after aluminum extrusion die design and the low efficiency of finding similar profile surface dies,a matching algorithm based on the structure characteristics of profile surface was proposed.According to the new profile surface image matching database of similar profile surface images,the corresponding die design was found for reference.It used YOLOv5 algorithm to effectively identify the structural features of the profile surface and automatically mark the structure position.By calculating the position,acreage,and type of the corresponding structure on the profile surface,the structure similarity was obtained,and the structure was matched one-to-one according to the idea of minimum cost and maximum flow to obtain the best matching result.
作者
周钢亮
秦昊
林卓胜
张昱
王培龙
ZHOU Gangliang;QIN Hao;LIN Zhuosheng;ZHANG Yu;WANG Peilong(Faculty of Intelligent Manufacturing,Wuyi University,Jiangmen 529020,China;Institute of Intelligent Manufacturing,Guangdong Academy of Sciences,Guangzhou 510075,China)
出处
《模具工业》
2024年第5期12-19,共8页
Die & Mould Industry
基金
铝型材挤压工艺过程管控系列工业软件研发及应用示范项目(2021B0101220006)。
关键词
铝型材截面
结构特征
图像匹配
机器学习
aluminum profile section
structural characteristics
image matching
machine learning
