The detection and alignment of polarization axis is a key technique in the applications of polarization maintaining fiber (PMF).Considering the fact that the detection accuracy of PMF ′s polarization axis using the transverse observation method is not high,this paper proposes an accurate detec tion method based on image processing of PMF′s end surface.Firstly,this method speeds up the recognition of PMF’s stress area and enhances the recognition robustness through the application of an auto-adaptive threshol d segmentation based on three characteristics of the contour of the stress area.Then the method improves the location accuracy of the center of PMF’s stress area by using an accurate location technology,which includes thre e steps that are extracting concentric circles corresponding to different gray values,fitting circles with interring method to exclude non edge points,and calculating the weighted mean of the centers of the fitting circles. Finally,we analyze the influence factors of polarization axis alignment and establish an error formula for polari zation axis alignment.The experimental results show that the recognition algorithm in this paper costs l ess time and has higher robustness to different light environment conditions than the traditional method of template m atching.Besides,both the simulation results and experiment results indicate that the detection algorithm in this paper can a chieve the accuracy less than ±0.1°.