Automated vision-based inspection technique using single-camera calibration for evaluating ISO dimensional specifications

Recently, artificial intelligence has been widely explored in the field of manufacturing. Machine vision, based on image processing, is one of the intelligent techniques used for on-machine dimensional inspection. Obtained point clouds are highly dependent on the configuration of the camera used for image acquisition. Thus, a calibration process is required to obtain a consistent point cloud for the evaluation of dimensional specifications. In this paper, a new machine vision-based camera calibration technique is proposed to improve the efficiency of on-machine dimensional inspection. The technique is based on the use of artifacts and a processing technique for point cloud resizing. The camera model parameters are determined through an inverse design optimization process. Specifications are then evaluated on the basis of the calibrated point clouds. The impact of camera positioning is investigated within this context.