Abstract: The self-driving Vehicles attitude positioning based on monocular vision has the advantages of simple structure and easy operation, but it is difficult to solve the problem of positioning accuracy and speed. In this paper, an improved method based on essential matrix decomposition for self-driving Vehicles attitude estimation is proposed. Firstly, according to the projection mapping relationship of the camera, the feature matching point pairs are extracted by the feature of small angle change between two consecutive frames, and the basic matrix is solved. An essential matrix decomposition algorithm that does not rely on singular value decomposition, and a method for determining the unique solution of the optimization, solves the real-time positioning problem of the self-driving Vehicles. Finally, the efficiency and accuracy of the algorithm are analyzed by simulation experiments. Compared with the existing algorithms, the experimental results show that the improved attitude estimation algorithm can effectively improve the operation speed. It has certain reference value for the real-time positioning research of self-driving Vehicles. Key words: monocular vision; pose estimation; essential matrix;

CICTP