Multi-Camera System with Advanced Image Correction Techniques for Industrial Applications (IMPACT 2024)

This study proposes a novel multi-camera calibration method specifically designed for industrial PCB (Printed Circuit Board) applications. The proposed method utilizes small targets to calibrate large industrial objects without requiring overlapping camera fields of view. It employs triangulation and other techniques to determine camera poses from partial images captured by the cameras. Image reconstruction techniques are then used to verify the camera poses. This approach does not necessitate additional sensors or projection equipment. Instead, it utilizes simulation software to set precise parameters, effectively reducing calibration errors and sensitivity. The method offers several advantages, including high accuracy, low cost, and simple structure, making it suitable for on-site camera calibration. By integrating this innovative multi-camera calibration method, the system provides a reliable and efficient solution for the precise calibration of cameras in PCB manufacturing and inspection processes. This enhances the overall quality control an d accuracy in the production of PCBs, ensuring high standards in the final product. Multi-camera systems hold significant importance in current industrial applications, particularly for scene reconstruction and camera pose estimation, which remain popular research topics. However, measuring large workpieces continues to be challenging. Our method offers an integrated approach using both virtual and real environments to estimate camera poses even when multiple cameras have non-overlapping fields of view (FOV).