Abstract: When the autonomous and remotely operated vehicle (ARV) performs close-range underwater docking, there exist some problems, such as poor performance of the positioning method and relatively large errors of the positioning information. For those problems, an underwater visual positioning algorithm based on the reflective tape scheme is designed. Firstly, a Q-learning based hybrid whale optimization algorithm (QHWOA) is proposed to improve convergence accuracy and convergence speed. With the reflective tape placed outside the interface of the TMS (tether management system) as the target, the docking image is segmented through Otsu's objective function optimized by QHWOA. Then, a keypoint extracting algorithm based on PCA (principal component analysis) dimensionality reduction is proposed to extract key contours and keypoints. Finally, SRPnP (a simple, robust and fast method for the perspective-n-point problem) algorithm is used to calculate relative pose of the ARV and the TMS. Through underwater docking experiment, the errors of key point extraction in pixel and visual positioning are calculated, which verify that the positioning errors meet the accuracy requirements of underwater docking. The algorithm can output effective positioning information during underwater docking, and guide ARV docking with TMS.