摘要
针对光照变化导致的材料缝隙焊接识别与定位效果不理想的问题,提出一种基于图像处理的缝隙焊接识别方法,通过图像特征提取、改进BP的缝隙焊接与非缝隙焊接图像分类、缝隙焊接图像识别定位等步骤,实现了缝隙焊接的识别。MATLAB仿真结果表明,基于图像处理的缝隙焊接识别方法可在不同光照条件有效识别定位材料的缝隙焊接,且有重叠识别的检测率达到91.5%,虚警率为7.2%,无重叠识别的检测率为86.3%,虚警率为4.3%。相较于传统的Ostu识别,这方法检测率更高,性能更好。
Aiming at the problem that the recognition and location effect of seam welding is not ideal due to the change of light,a new method of seam welding recognition based on image processing is proposed.Through image feature extraction,gap welding and non seam welding image classification based on improved BP network,and seam welding image recognition and location,the seam welding recognition is realized.Finally,the validity of the identification method is verified by simulation in MATLAB software.The results show that the method of seam welding recognition based on image processing can effectively identify the location seam welding under different lighting conditions.The detection rate of overlap recognition is 91.5%,the false alarm rate is 7.2%,the detection rate of non overlapping recognition is 86.3%,and the false alarm rate is 4.3%.Compared with traditional recognition method Ostu recognition,the detection rate and performance of this method are better.
作者
程晓钰
CHENG Xiaoyu(School of Mathematics and Computer Science,Yunnan University for Nationalities,Kunming 650500,China)
出处
《粘接》
CAS
2022年第1期99-105,共7页
Adhesion
关键词
图像识别
图像处理
BP神经网络
LM算法
Image recognition
image processing
BP neural network
LM algorithm
