SMD Classification for Automated Optical Inspection Machine Using Convolution Neural Network

Because different inspection items are applied to different devices mounted on a printed circuit board (PCB), the types of such devices need to be identified for an automatic PCB inspection. This study proposes a device classification method using a convolution neural network (CNN), which is a deep learning technique. CNNs perform well in terms of image classification, and because many upsampling methods have been recently proposed, such an approach is widely used for segmentation problems. The classification method proposed in the present study uses a CNN to extract and classify only the device regions from device images obtained from a PCB.