在未知的环境中，自主水下航行器（AUV）运动未知的情况下，采用单目视觉不能对路标进行完全定位，存在尺度漂移。针对这种情况，研究了融合声纳测距信息的AUV定位方法，建立了AUV的运动模型，路标的初始化模型及相机的测量模型，采用扩展卡尔曼滤波对AUV位置进行估计；为了提高定位算法的实时性，提出了基于预测的特征匹配与误匹配点去除算法，既提高了匹配的效率，又提高了匹配的正确率。实验结果表明，该算法定位较准确，实现了AUV的三维定位。

In unknown environments, when AUV motions are unknown, the AUV and landmarks can not be completely located with monocular vision due to scale drift. For this situation, the AUV location method with the fusion of sonar distancemeasuring information was studied, and the models of AUV motion, landmark initialization and camera measurement were built, then the EKF filter was used to estimate the AUV position. In order to improve the realtime performance of the localization algorithm, the feather matching and error matching removing algorithm based on forecast was proposed. The experimental results show that the location algorithm realizes the AUV 3D localization accurately.