基于深度学习的PCB芯片极性检测算法

PCB Chip Polarity Detection Algorithm Based on Deep Learning

印刷电路板芯片极性检测是印刷电路板缺陷检测的重要组成部分。针对传统印刷电路板芯片尤其是小型芯片的极性检测算法准确性的不足,提出一种基于深度学习的电路板芯片极性检测方法。该方法首先使用模板匹配对不同电路板上不同型号芯片进行识别定位;其次,对电路板参考图像与待测图像进行配准、灰度化、二值化,并对两幅图像进行差分处理得到差分图;最后,结合芯片识别定位结果提取差分图中的芯片区域,并采用提出的卷积神经网络实现对芯片的极性检测。实验表明,所提方法比传统方法具有更高的检测准确率,准确率可达98.26%,满足工业检测精度需求。

The polarity detection of chips on printed circuit board is an important part of circuit board defect detection. Aiming at the problem that the accuracy of the polarity detection methods for traditional chips, especailly tiny chips, needs to be improved, a printed circuit board chip polarity detection method based on deep learning is proposed. Firstly, template matching is used to identify and locate different types of chips on different circuit boards. Secondly, the testing image is registrated to the reference image, grayscale and binarization are operated on both images, then two images are subtracted to get the difference map. Finally, the chip area in the difference map is extracted, and the polarity detection of the chip is realized by using the proposed convolutional neural network. Experiments show that the method has higher detection accuracy than the traditional methods, and the accuracy of proposed method achieves 98.26 percent, reaching the requirements of industrial inspection accuracy.