Image Denoising of Printed Circuit Boards using Conditional Generative Adversarial Network
Hsien-I Lin, Pedro Menendez · 2019
The objective of the paper is to detect whether a needle marker occurs in the pad area of a printed circuit board from images. Since there exist many irrelevant noises including other small circuit pads, it becomes difficult to find the pad area and detect the needle marker. In our proposed approach, we present a denoising method to find the pad area of a printed circuit board from images. The proposed method adopts a conditional generative adversarial network (CGAN) to denoise the images and has better results than traditional image processing techniques. The method proceeds in three steps: firstly, the input images are classified by “single pad” and “multiple pad” using a convolutional neural network (CNN). Secondly, the image with a single pad is denoised by a CGAN model. Thirdly, the image is cropped by the pad region and used to find the needle marker by pattern matching. The effectiveness of the proposed method is validated with the experimental results and shows that the needle marker is accurately detected.