摘要
针对当前焊接缺陷识别算法难以检测出低对比度噪声焊接缺陷,以及难以满足实时检测需求等不足,文章提出了加权中值各向扩散模型耦合感兴趣区域的放射图像焊接缺陷实时检测技术。基于感兴趣区域原理,确定出焊接缺陷位置,降低复杂度,并可显著提高检测精度;引入中值滤波,建立焊接缺陷区域图像梯度计算模型;再设置门槛系数K,嵌入非线性S函数,设计新的边缘停止规则;并融合锐化算子,设计了加权中值各向扩散模型,完成焊接缺陷检测。文中设计的加权中值各向扩散模型能够根据局部梯度幅值,自适应地完成增强、锐化焊接缺陷以及平滑放射图像背景。仿真结果显示:与当前焊接缺陷检测技术相比,该方法能够更好地增强焊接缺陷细节,可有效区分图像背景与焊接缺陷特征;且拥有更高的缺陷检测精度与效率。
In order to Solve these defects such as can not detection the weld defects of low contrast and meet the requirements of real time in current weld defects detection algorithms, the weld defects real time detec-tion technology based on weight median anisotropic diffusion model and region of interest principle was pro-posed in this paper. The location of weld defects was determined by region of interest principle to reduce the complexity and improve the detection precision;then the image gradient computation model of weld defects was established by introducing the median filter. The new edge stopping rule was designed by embedding the non-linear S function and setting the threshold coefficient K. And the weighted median anisotropic diffusion model was designed by integrating the sharpening operator to finish the weld defects detection. This model can adaptively enhancement, sharpen weld defects and smooth the background of radiation image according to the local gradient magnitude. Simulation results showed that：this defects detection technology can better enhance the weld defects details to efficiently distinguish the image background and weld defects Features;and the defects detection precise and efficiency was higher.
出处
《组合机床与自动化加工技术》
北大核心
2015年第9期86-90,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
江西省自然科学基金项目(2011GZW0049)
江西省教育厅项目(GJJ12632)
关键词
加权中值各向扩散
感兴趣区域
非线性S函数
weight median anisotropic diffusion model
region of interest
non-linear S function
