基于亚像素的PCB表面质量检测

Surface Quality Inspection of PCB Based on Subpix

为提高PCB的测量精度及缺陷识别率,针对不同类型的边缘,设计了亚像素边缘定位拟合模型,实现了PCB光板上导线及焊盘的亚像素级边缘检测。在此基础上,设计了神经网络分类器,提取PCB光板上导线及焊盘的特征作为输入,通过网络训练,构建了用于缺陷检测的MLP神经网络模型。测试实验表明,亚像素边缘检测方法可使PCB光板图像的测量精度进一步提高;神经网络分类器可有效识别和分类PCB光板表面常见的缺陷。

In order to improve the measuring accuracy of electronic circuit board and defect recognition rate,a subpixeledge location fitting model was designed according to different types of edge,which implemented subpixeledge detection of wires and solder on bare PCB board.Further more,neural network classifier was designed,and the characteristics of the wire and welding disk on the PCB plate were extracted as input,and the MLP neural network model for defect detection was constructed by training.The test results show that the measurement accuracy of PCB image is further improved by subpixeledge detection method.And neural network classifier can effectively realize the recognition and classification of common defects on PCB surface.