Polarization axis detection method of imaging defect fiber images based on neighborhood adaptive adjustment

Aiming to address the challenge of accurately identifying the polarization axis in polarization-maintaining fiber images with imaging defects, such as abnormal local brightness and low contrast between the fiber end face and the stress region, a detection method for the polarization axis of imaging-defective fiber images based on neighborhood-adaptive adjustment was proposed. Initially, the method adjusted the anomalous pixel values of the fiber end face by constructing an adaptive global threshold. Subsequently, a dynamic local threshold was formulated to correct the pixel values in the transition zone between the fiber end face and the stress region. Finally, sub-pixel interpolation was employed to extract the contour of the segmented stress area, thereby achieving sub-pixel-level precise positioning of the stress area boundary and the polarization axis of the defective optical fiber image. Experimental results demonstrated that compared with the Canny-Otsu algorithm, Canny-MS-Scharr algorithm, Canny-EC algorithm and Canny-SFNI algorithm, the proposed method reduced the detection error in the stress region by over 80%, while maintaining the detection deviation of the polarization angle under Gaussian and salt-and-pepper noise influences within 0.3°.