基于神经网络的印制线路板图件识别研究

Research on Recognition of Diagram Pieces Rooting in PCB Based on Neural Network

对基于人工神经网络的印制线路板(PCB)图件识别进行了研究。识别过程主要分为两个阶段:图像的预处理和模式识别。图像预处理阶段运用数字图像处理技术,从原始PCB图中提取出待识别的图件对象;在模式识别阶段,采用BP神经网络模型对PCB图件图像进行识别,为降低网络规模,经深入分析比较,提出了一种行、列扫描提取图件图像特征向量的算法,并对标准BP学习算法进行了改进,获得了较好的效果。

The paper research the application of artificial neural network in the recognition of diagram pieces rooting in PCB. The recognition makes up of two phases： image pretreatment and recognition. In the first phase, we distill the diagram pieces from the original image of PCB by the technique of digital image process; In the second phase, recognizing the diagram pieces of PCB using the BP neural network, in order to lower the scale of the network, we put forward a algorithm extracting the feature vector of diagram pieces by the row or diagram column scanning, in addition, we mend the standard algorithm in the study course of BP neural network, and achieving better effect.