融合显著性检测的V型焊缝激光条纹提取
Laser Stripe Extraction of V-Shaped Weld Seam Based on Visual Attention Mechanism
任紫芸 1任红格 2李福进1
作者信息
- 1. 华北理工大学电气工程学院,河北 唐山 063210
- 2. 天津城建大学控制与机械工程学院,天津 300384
- 折叠
摘要
针对在背景亮度不均的情况下提取的V型焊缝坡口激光带准确度低的问题,提出一种基于视觉注意机制融合颜色特征和方向特征的方法.首先,使用FT(Frequency-Tuned)显著性检测算法获取焊缝图像中激光带的颜色显著图.然后,利用Gabor滤波器得到激光带的单方向特征图,对各方向图阈值分割消除噪声后等比例加权融合得到方向显著图.使用多层元胞自动机融合颜色显著图和细化后的方向显著图,得到最终的焊缝激光带条纹.实验结果表明该方法简单有效,效果较好,准确度高,为后续工作打下基础.
Abstract
A method of fusing laser stripe color features and directional features is proposed for the problem of low accuracy in ex-tracting laser stripe of V-shaped weld bevels from a background with uneven brightness.The color features of the laser stripe in the weld image are firstly obtained by FT(Frequency-Tuned)algorithm.And the single directional feature maps of the laser stripe are obtained by using Gabor filtering.Then the noise is eliminated by threshold segmentation of each directional map.The direc-tional feature map is obtained by equal proportional weighted fusion.A multilayer cellular automaton is used to fuse the color fea-ture map and the refined direction feature map to obtain the weld laser stripes.The experimental results show that the method is simple and effective with good results and high accuracy,and it lays the foundation for the subsequent work.
关键词
激光条纹提取/FT算法/Gabor滤波/阈值分割/图像细化/元胞自动机Key words
Laser Stripe Extraction/FT Algorithm/Gabor Filtering/Threshold Segmentation/Image Refine-ment/Cellular Automata引用本文复制引用
基金项目
河北省自然科学基金项目(F2018209289)
国家自然科学基金项目(61203343)
出版年
2024
NETL
NSTL
维普
万方数据