摘要
为了实现对产品表面缺陷的快速自动识别,提出在先验已知各目标区域的最小待识别结构尺寸及空间畸变的条件下,采用最小二乘法拟合得出采样步长变化的函数关系曲面及其曲面对应的公式表达。将各待检区域的因变量带入公式计算可得到采样步长,在满足采样定理的前提下,基于变步长采样机制获取标准图像序列库;其次,通过LTP算法寻找被检产品的图像在标准库中的最优位置;最后,通过相关度判别各区域有无缺陷。实验表明在实际工业检测应用中,工程技术人员可以利用采样步长与最小待识别结构尺寸及空间畸变的函数关系确定采样步长,建立变步长采样机制。
In order to realize the rapid automatic identification of product surface defects, under the condition of known the minimum resolution and spatial distortion of target regions, using the least squares fitting concluded that the function relation surface of sampling step change and its corresponding formula. The dependent variable of each quarantine regions into the formula to calculate that sampling step can be obtained, when meet sampling theorem based on variable step sampling mechanism will get standard library image sequences. Secondly, by the LTP algorithm to find that tested product images in the optimal location of the standard library. Finally, through the correlation to determine whether defect regions. Experiments show that in actual industrial detection applications,engineering and technical personnel can establish variable-step mechanism according to the function relation of sampling step and minimum resolution and space distortion.
作者
郭静
韩跃平
李会鸽
Guo Jing;Han Yueping;Li Huige(School of Information and Communication Engineering,North University of China,Taiyuan 030051,China)
出处
《科技通报》
北大核心
2017年第2期129-132,200,共5页
Bulletin of Science and Technology
基金
国家自然科学基金(61171178)
山西省自然科学基金(2012011010-3)
2012年山西省高等学校优秀学术带头人支持计划
关键词
变步长采样机制
序列图像
缺陷检测
拟合
variable-step mechanism
image sequences
defect detection
fitting
