Method for improving monocular vision initialization stability

The invention relates to a method for improving monocular vision initialization stability. Firstly, a Harris feature point detection method and LSD are used for extracting point features and line features between each frame of image, feature tracking is carried out between every two frames of images, and corresponding feature pairs between every two frames of images are obtained. And an essential matrix is calculated by using the feature point pairs, the essential matrix is optimized by maximizing the projection overlapping length of the line features, and then the essential matrix is decomposed to calculate the pose to obtain a stable initialization result. The monocular vision SLAM initialization robustness and the positioning precision can be effectively improved by the monocular vision SLAM initialization method. 本发明涉及一种提高单目视觉初始化稳定性的方法。首先在每帧帧图像间利用Harris特征点检测方法以及LSD提取点特征和线特征，在每两帧图像间进行特征的跟踪，获取每两帧图像间对应的特征对。再利用特征点对解算本质矩阵，然后通过最大化线特征的投影重叠长度优化本质矩阵，之后分解本质矩阵解算位姿，获得稳定的初始化结果。本发明能有效提高单目视觉SLAM初始化的鲁棒性以及定位精度。