基于神经网络的PCB焊点检测方法研究

随着印刷电路板（PCB）组装技术向高密度化和“零缺陷”方向发展，市场对自动光学检测系统（AOI）的要求也向高准确率、智能化发展。针对目前AOI在焊点检测时，容易出现缺陷误报和漏报，以及智能化程度不高的情况，作者提出了一种基于神经网络（ANN）的检测方法。首先，采用了一种基于熵的多阈值自动图像分割方法；然后，定义了焊点图像的一系列特征，并通过实验对特征进行选择；最后，建立了用于进行焊点分类的BP神经网络。实验证明，基于神经网络的焊点图像检测方法具有较高的准确率。

Solder Joint Inspection Method Based on Artificial Neuron Network

As component sizes in electronics get smaller and board densities become more compact, the need for automatic inspection in electronic manufacturing is becoming a must. The automatic optical inspection (AOI) system is demanded more precisely and intelligent. To overcome the disadvantage of misalarm, a method based on artificial neuron network was proposed. Firstly, a multi-thresholding algorithm based on entropy was adopted. Then, a series of features of solder joints were defined and the principal features were selected. Finally, the artificial neuron network was established for classification the solder joints. The performance of the method was verified by the experiment.