In order to improve the accuracy of feature based stereo visual localiz ation algorithm under low-light environments,we propose a novel low-light ima ge enhancement algorithm based on online estimation for visual simultaneous locali zation and mapping (simultaneous localization and mapping,SLAM).By online esti mating the image brightness value,the corresponding parameters of the image enh ancement algorithm are updated in real time.Therefore,it solves the problem of inapplicability of standard parameter-fixed image enhancement algorithm,which performs poorly in the case of very bright or dark environments.First,these fa ctors which affect the localization accuracy are explored in the ORB-SLAM2syst e m,and their corresponding parameters are updated in real time by online paramet er estimation method.Then,the low-light image enhancement (LIME) algorithm is applied to improve the image quality of low-light e nvironments.Finally,the image feature extraction is performed on the enhanced image effectively,which improves feature matching accuracy,and it is also bene ficial for improving the localization accuracy.The proposed method is extensive ly validated on the public EuRoC datasets and compared with the currently widely used ORB-SLAM2algorithm.The experimental results effectively verify that the proposed visual SLAM system achieves better localization accuracy and robustness than the others.