摘要
提出一种基于稀疏求解的输油管道焊缝图像缺陷检测算法,通过求解稀疏系数完成识别过程,避免了图像的分割与特征值的计算。将现场提取的缺陷和噪声图像作为样本,根据压缩感知理论,首先通过学习从样本中获得字典,利用相关性最小的原则确定字典矩阵数量,然后构建稀疏模型并采用正交匹配追踪算法求解该模型,最后通过求解得出的系数组合判断图像类型。实验表明:该方法能够实现图像缺陷的准确识别。
A sparse solution-based image defect detection algorithm for oil pipeline welds is proposed.The recognition process is completed by solving the sparse coefficient,which avoids image segmentation and feature value calculation.Taking the defect and noise images extracted from the scene as samples,ac-cording to the compressed sensing theory,first obtain dictionaries from the samples through learning,use the principle of least correlation to determine the number of dictionary matrices,then construct a sparse model and use an orthogonal matching pursuit algorithm to solve the model,And finally judge the image type through the coefficient combination obtained by the solution.
作者
王侦倪
刘怡婷
李阳
WANG Zhen-ni;LIU Yi-ting;LI Yang(Research Institute of Shaanxi Yanchang Petroleum(Group)Co.,Ltd.,710000,China)
出处
《内蒙古石油化工》
CAS
2021年第7期4-5,110,共3页
Inner Mongolia Petrochemical Industry
关键词
图像处理
稀疏建模
贪心算法
缺陷识别
image processing
sparse modeling
greedy algorithm
defect recognition