基于INS/前视声纳组合的输水隧洞AUV巡检定位方法

当潜航器(AUV)进行输水隧洞巡检时,多普勒测速仪(DVL)声波会对前视声纳图像产生干扰,针对这一问题提出了一种从AUV上卸载DVL,直接从前视声纳图像中提取栽体速度,而后和惯性测量单元(IMU)进行组合导航的方法.该方法根据前视声纳成像原理,建立起每相邻两帧声纳图像间载体的相对位移与配对特征点的图像坐标之间的联系,针对水底为局部平坦区域的情况,对特征点仰角进行了估算.通过惯导速度约束和随机抽样一致(RANSAC)算法,剔除误差较大的配对点,然后利用提取出的载体速度作为观测量进行卡尔曼滤波.经实际数据测试,惯性导航系统(INS)/前视声纳组合的总体性能和INS/DVL组合非常接近,以输水隧洞内的接缝线作为定位基准,INS/前视声纳组合导航在沿隧洞方向上的最大相对定位误差小于行程的1％.

Positioning Method of AUV for Water Conveyance Tunnel Inspection Based on INS/Forward-looking Sonar Integration

When an AUV is carrying out routine inspection in a water conveyance tunnel, the sonic wave of DVL will cause interference to the forward-looking sonar image. To solve this problem, a method of unloading DVL from AUV, directly extracting the vehicle velocity from the forward-looking sonar image, and then making it combined with an IMU is proposed. According to the principle of forward-looking sonar imaging, the method establishes the relationship between the relative displacement of the vehicle between two adjacent sonar images and the image coordinates of the paired feature points, and estimates the elevation of feature points for the situation in which the water bottom is locally flat. By using the inertial velocity constraint and RANSAC algorithm, the matched points with large error are eliminated. Finally, the extracted vehicle velocity is used as observed quantity for the Kalman filter. In real world experiment, the overall performance of the INS/forward-looking sonar integration is very close to that of the INS/DVL integration. By taking the seam lines in the water conveyance tunnel as positioning reference, the maximum positioning error of the INS/forward-looking sonar integration in the direction of tunnel is less than 1% of the total distance travelled.