Metaheuristic Camera Calibration for Optical Tomographic Imaging in Industrial Environments

Providing 3D tomographic imaging in industrial combustion processes can improve process control, optimize fuel consumption and reduce harmful products. The application of this imaging technique that features extreme parameters in such environments requires the ability to reliably calibrate the optical setup. We introduce a camera calibration method based on a genetic algorithm that leverages corners detected from 3D ChArUco calibration target images. For corner detection in images of ordinary checkerboard targets, a convolutional neural network algorithm was developed. Our genetic algorithm estimates intrinsic parameters of a pinhole camera combined with a nonlinear lens distortion model and simultaneously optimizes its distortion and inverse distortion coefficients. Our results were compared well with existing calibration toolboxes and libraries based on Zhang’s well-established method. Additionally, to validate the accuracy of our calibration algorithm, it was applied to a tomographic chemiluminescence measurement of a turbulent flame using an array of cameras equipped with wide-angle lenses.