摘要
在零件加工过程中,能够及时的发现加工工件的表面缺陷,可以保证加工质量。针对由于刀痕影响工件表面缺陷不易检测的问题,提出一种基于谱残差视觉显著模型的工件表面缺陷检测算法。首先,对工件表面图像预处理,使用同态滤波方法消除工件表面光照不均匀的影响;然后提出改进的谱残差视觉显著模型,大致定位缺陷;随后使用视觉显著性和超像素分割相结合的算法来进一步确定缺陷的位置;最后使用形态学运算,得到最终的检测结果。在现场采集图片库上客观实验评估表明,该算法具有很高的准确率且检测速度快。
In the automatic processing of parts, testing the surface defects of those processing parts in time can ensure their quality. Aiming at the problem that the surface defects of machined parts are not easy to be detected due to the influence of tool marks, a surface detection algorithm for machined parts based on spectral residual visual saliency model is proposed. Firstly, the surface image of the part is preprocessed, and using the homomorphic filtering method to remove the influence of uneven illumination on the surface of the part;then an improved spectral residual visually significant model is proposed to locate defects;then the algorithm of combining the visual saliency and superpixel segmentation is used to further determine the location of defects;Finally, the morphological operation is used to obtain the final detection result. The objective experimental evaluation on the on-site collection picture library shows that the algorithm has high accuracy and fast detection speed.
作者
许杨
刘丽冰
杨泽青
黄凤荣
XU Yang;LIU Li-bing;YANG Ze-qing;HUANG Feng-rong(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300131,China)
出处
《组合机床与自动化加工技术》
北大核心
2019年第10期102-105,110,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
河北省科技计划项目(16211803D)
天津市自然科学基金一般项目(16JCYBJC19100)
河北省自然科学基金和重点基础研究专项(E2017202294)
关键词
缺陷检测
同态滤波
谱残差
超像素
defect detection
homomorphic filtering
spectral residual
superpixel
