Unsupervised polarization imaging system polarization detection direction self-calibration method
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING INST OF TECH
- Filing Date
- 2025-12-18
- Publication Date
- 2026-05-26
AI Technical Summary
Existing polarization imaging system calibration methods have stringent requirements. Supervised optimization algorithms require the true value of the polarization angle and are prone to calibration errors, while unsupervised optimization algorithms are only applicable to 4-channel systems and are difficult to achieve self-calibration of 3-channel and above systems when the true value of the polarization angle is unknown.
An unsupervised optimization algorithm is used to image scenes with unknown polarization information through a polarization imaging system, acquiring multiple sets of intensity images at different angles. Combining the mechanism of polarization degree stability, a gradient descent algorithm is used to adjust the polarization detection direction of the polarization system, and an error model is built to achieve self-calibration. This method is suitable for systems with 3 channels or more.
It achieves fast and accurate self-calibration of polarization system polarization direction under unknown true polarization angle conditions, with a wider range of applications, faster calculation speed, and stronger robustness, and is suitable for 3-channel and above systems.
Smart Images

Figure CN122084544A_ABST