Proceedings of the
9th International Conference of Asian Society for Precision Engineering and Nanotechnology (ASPEN2022)
15 – 18 November 2022, Singapore
doi:10.3850/978-981-18-6021-8_OR-11-0121
Optimization Strategy of Camera Calibration for Close Range Photogrammetry
1Laboratoire Commun de Métrologie (LCM), Laboratoire National de Métrologie et ďEssais (LNE), 1 Rue Gaston Boissier, 75015, Paris, France
2Université Paris-Saclay, ENS Paris-Saclay, LURPA, Université Sorbonne Paris Nord, 91190, Gif-sur-Yvette, France
3College of Engineering, Prince Sattam bin Abdulaziz University (PSAU), Department of Mechanical Engineering, Alkharj 16273, Saudi Arabia
ABSTRACT
Camera-based scanning systems are increasingly used for 3D reconstruction and inspection due to their high-density measurements in real time. Preliminary to 3D scanning, a calibration process is required to estimate the internal and external parameters of the camera model, also known as intrinsic and extrinsic parameters. The calibration process estimates the model parameters while establishing the relationship between 3D points and their projection on the image plane. This estimation is achieved by taking images of a pattern with known dimensions from multiple camera poses.
The accuracy of the calibration directly affects the accuracy of 3D scanning. The estimated parameter's accuracy depends on the selected camera poses during the calibration process. Therefore, optimal pose selections have been investigated to improve the calibration. The proposed optimal pose selection is based on polynomial regression and particle swarm optimization. Polynomial regression is applied to a dataset containing reprojection errors, to quantify the difference between the estimated and measured pattern projections, and their corresponding camera poses. Particle swarm optimization under linear and non-linear constraints allows to find the optimal poses corresponding to minimum reprojection errors in the polynomial. The selected optimal poses are then used for the calibration.
Simulated and experimental calibrations have been performed while evaluating the accuracy of the estimated intrinsic and extrinsic parameters. The proposed optimization strategy provides accurate estimations in comparison to recently reported works named Calibration Wizard and Efficient Pose Selection. Errors in focal length parameters estimated by the proposed approach are reduced by 20% compared to recent estimation errors, and reductions are higher than 70% for all other parameters.
Keywords: Camera calibration, Polynomial regression, Particle swarm optimization, Photogrammetry, Optical measurement
1Laboratoire Commun de Métrologie (LCM), Laboratoire National de Métrologie et ďEssais (LNE), 1 Rue Gaston Boissier, 75015, Paris, France
2Université Paris-Saclay, ENS Paris-Saclay, LURPA, Université Sorbonne Paris Nord, 91190, Gif-sur-Yvette, France
3College of Engineering, Prince Sattam bin Abdulaziz University (PSAU), Department of Mechanical Engineering, Alkharj 16273, Saudi Arabia
ABSTRACT
Camera-based scanning systems are increasingly used for 3D reconstruction and inspection due to their high-density measurements in real time. Preliminary to 3D scanning, a calibration process is required to estimate the internal and external parameters of the camera model, also known as intrinsic and extrinsic parameters. The calibration process estimates the model parameters while establishing the relationship between 3D points and their projection on the image plane. This estimation is achieved by taking images of a pattern with known dimensions from multiple camera poses.
The accuracy of the calibration directly affects the accuracy of 3D scanning. The estimated parameter's accuracy depends on the selected camera poses during the calibration process. Therefore, optimal pose selections have been investigated to improve the calibration. The proposed optimal pose selection is based on polynomial regression and particle swarm optimization. Polynomial regression is applied to a dataset containing reprojection errors, to quantify the difference between the estimated and measured pattern projections, and their corresponding camera poses. Particle swarm optimization under linear and non-linear constraints allows to find the optimal poses corresponding to minimum reprojection errors in the polynomial. The selected optimal poses are then used for the calibration.
Simulated and experimental calibrations have been performed while evaluating the accuracy of the estimated intrinsic and extrinsic parameters. The proposed optimization strategy provides accurate estimations in comparison to recently reported works named Calibration Wizard and Efficient Pose Selection. Errors in focal length parameters estimated by the proposed approach are reduced by 20% compared to recent estimation errors, and reductions are higher than 70% for all other parameters.
Keywords: Camera calibration, Polynomial regression, Particle swarm optimization, Photogrammetry, Optical measurement
