一种地标点修正的高精度双目视觉导航方法

自主导航技术是无人机、智能机器人以及智能车辆等技术领域中的一项关键技术。为了克服传统视觉导航方法存在的不足，以摄像机自运动估计为理论基础，提出了一种使用地标点修正的高精度双目视觉导航方法。该方法从连续图像序列的帧间变化中提取摄像机自运动信息，再通过速度、角速度积分求得载体的位置姿态等参数。同时，通过引入地标点提供的绝对定位信息，进一步提高了长时间导航的精度。将地标点绝对定位与双目视觉相对定位相结合，并对双目视觉量测噪声建模和解相关，抑制了单纯使用双目视觉长时间导航时误差的积累发散。仿真结果表明此方法具有精度高、自主性强、导航信息完备等优点。

Landmark fixed high-precision binocular visual navigation method

Autonomous navigation is a crucial technology in the field of unmanned aerial vehicle (UAV), intelligent robots and smartcars. To overcome shortages in traditional visual navigation methods, a landmark fixed high-precision binocular visual autonomous navigation method, based on the camera egomotion estimation theory, was put forward. In this method, camera's egomotion parameters, velocity and angular velocity, were measured from the variance between frames of the continuing image sequence, then the position and attitude was calculated by accumulating the velocity and angular velocity. Moreover, by bringing in the absolute positioning information provided by the landmarks, the performance of the method in long time navigation was improved. Combining the absolute positioning information from the landmarks and the relative navigation information from the binocular vision, as well as modeling and decoupling the measurement noise of the binocular vision, the accumulation and increasing of the navigation error was suppressed. Simulation result shows that this method has the advantage of high precision, autonomy and completeness of navigation information.