摘要
针对现阶段基于图像检测的板形识别在缺陷的检出率普遍较低以及检测速度慢等问题,文中提出了一种基于计算机图像处理的板形识别系统,其通过直方图均衡化与高帽变换对初始图像进行处理,并通过边缘检测算法提取轮廓,然后利用BP神经网络分类器进行缺陷识别与分类。其在实验及实际工业生产中,均具有较高的识别率,可达到约90%,且还具有较好的板形识别效果。
For this stage flatness recognition based on image detection in defect detection rate is generally low and slow detection speed and other issues, this paper presents a computer-based image processing board shape recognition system,which by histogram equalization and hat transform the initial image processing, and extracts contour through the edge detection algorithm, and then use BP neural network classifier defect recognition and classification. Its experiments and actual industrial production has high recognition rate, were about 90%, with good flatness recognition effect.
出处
《电子设计工程》
2016年第21期177-179,183,共4页
Electronic Design Engineering
关键词
计算机图像处理
板形识别
边缘检测
BP神经网络分类器
computer image processing
plate-profile recognition
edge detection
BP neural network classifier
