Improving Photogrammetry Instrument Performance through Camera Calibration for Precision Digital Manufacturing

Abstract

Serial-axis robots are being used in precision digital manufacturing for more intensive applications than previously expected. These industrial robots can show global inaccuracy and positional drift errors over time, which affects the quality of manufactured parts. Photogrammetry is the use of multiple cameras to measure objects in 3D environments, and in this study, camera calibration of a photogrammetry tool consisting of a stereo pair of infrared cameras is investigated. Accurate camera calibration is essential for precise measurement and monitoring of industrial robot positioning in 6 degrees of freedom (6DoF) by a photogrammetry tool. Possible areas of improvement in the camera calibration of the photogrammetric system include the calibration routine and the distortion model accuracy, which are explored in this paper. Variations of the calibration routine are tested and compared to the original routine. Routines with more images and even spacing through the node measurement volume are shown to have more repeatable parameter values. Wide-angle lenses are used to minimize setup volume; however, this results in large amounts of distortion towards the image edges. Improved distortion parameters could result in an increased useable image area. A further in-depth investigation is conducted into the accuracy of the distortion model within the camera calibration, which is found to overcorrect for distortion. Conclusions from these investigations can be used to improve camera calibration for the photogrammetric sensor network, ultimately improving the accuracy of the measurement system and the digital manufacturing platform.