PCBShot: An Assisted Image Acquisition Method for PCB Damage Detection With Mobile Devices

Identifying damages in Printed Circuit Boards is a critical task for quality assurance and repair inspection workflows. Image processing mobile applications, with embedded deep learning, assist technicians in detecting damages in this task, increasing accuracy and agility. However, the performance of such applications is highly dependent on the ability of the user in taking adequate photos. We propose an automatic capture method named PCBShot, that assists users of mobile applications of PCB damage detection to take better photos, enhancing the detection performance. Our method uses classical image processing algorithms to detect if a target PCB is inside a virtual guideline, ensuring that the position and distance are appropriate. Then, a photo is automatically captured, the background is cropped and the image is sliced into four quadrants for resolution preservation. The damage detection is performed in the slices. We evaluate our method through a real-life mobile application used in repair centers of an electronics manufacturer, comparing the detection performance with the manual image acquisition, without further assistance. Our results show that our method largely surpasses the manual acquisition, as it allows the capture of higher-quality images due to framing assistance with image processing methods, eliminating noisy backgrounds and preserving resolution.