TY - GEN
T1 - A Method for Radar-Camera Automatic Extrinsic Calibration Based on Improved Homography Estimation
AU - Liu, Xiaoyang
AU - Lu, Bohan
AU - Fu, Dongyi
AU - Xu, Yong
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - For the extrinsic parameter calibration problem of millimeter-wave radar and camera in the field of intelligent transportation monitoring, this study proposes an improved automatic calibration method by virtue of the calculation of homography matrix. The main improvements include: introducing a cross-modality target matching strategy based on position and density morphological similarity to enhance the accuracy of point pair matching; integrating the Huber loss function with the Levenberg-Marquardt algorithm to build a robust optimization framework, which enhances optimization stability and accuracy; designing a dynamic error compensation mechanism to improve the algorithm of adaptability to complex environments. Extensive experiments have verified that the improved algorithm significantly outperforms classic DLT and RANSAC algorithm in calibration accuracy. The effectiveness of each module has also been validated by ablation experiments.
AB - For the extrinsic parameter calibration problem of millimeter-wave radar and camera in the field of intelligent transportation monitoring, this study proposes an improved automatic calibration method by virtue of the calculation of homography matrix. The main improvements include: introducing a cross-modality target matching strategy based on position and density morphological similarity to enhance the accuracy of point pair matching; integrating the Huber loss function with the Levenberg-Marquardt algorithm to build a robust optimization framework, which enhances optimization stability and accuracy; designing a dynamic error compensation mechanism to improve the algorithm of adaptability to complex environments. Extensive experiments have verified that the improved algorithm significantly outperforms classic DLT and RANSAC algorithm in calibration accuracy. The effectiveness of each module has also been validated by ablation experiments.
KW - cross-modality strategy
KW - dynamic error compensation
KW - extrinsic calibration
KW - homography matrix
KW - intelligent transportation systems
UR - https://www.scopus.com/pages/publications/105043443706
U2 - 10.1109/ACAIT67930.2025.11522110
DO - 10.1109/ACAIT67930.2025.11522110
M3 - 会议稿件
AN - SCOPUS:105043443706
T3 - Proceedings of 2025 9th Asian Conference on Artificial Intelligence Technology, ACAIT 2025
SP - 105
EP - 114
BT - Proceedings of 2025 9th Asian Conference on Artificial Intelligence Technology, ACAIT 2025
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 9th Asian Conference on Artificial Intelligence Technology, ACAIT 2025
Y2 - 12 September 2025 through 14 September 2025
ER -