基于视觉补充的水稻插秧机多传感器组合定位研究

Multi-sensor integrated positioning of rice transplanter based on visual supplementation

为改善基于GNSS/INS组合定位的水稻插秧机在遇到遮挡、电磁干扰、传感器失效等情况时的导航效果,在原有GNSS/INS组合定位的基础上,提出一种视觉导航系统(vision navigation system,VNS)补充的水稻插秧机多传感器组合定位方法。首先设计改进的Otsu法和改进的Hough变换算法用于视觉定位信息提取,并构建插秧机和相机坐标系关系方程以求解位姿值;然后采用具有容错功能的联邦卡尔曼滤波算法将VNS输出的定位信息和GNSS、INS输出的定位信息进行融合;最后分别在水泥地和水田进行试验。结果显示,空旷水泥地场景下,GNSS/INS/VNS组合定位和GNSS/INS组合定位精度相近,而在遮挡水泥地场景下,GNSS/INS/VNS组合定位解算出的位置误差和航向误差的平均值分别为1.77 cm和0.99°,相较于GNSS/INS组合定位方法分别提高46.8%和61.5%;水田试验中,经过视觉补充后导航系统的横向偏差和航向偏差平均值分别降低45.7%和67.9%,横向偏差平均值为1.97cm,航向偏差平均值为0.49°。试验结果表明,基于视觉补充的多传感器组合定位方法能有效降低导航系统的定位误差和跟踪偏差,满足插秧机自动驾驶作业的要求。

A multi-sensor integrated positioning method for a rice transplanter supplemented with vi-sion navigation system based on the GNSS/INS integrated positioning was proposed to improve the naviga-tion effect of rice transplanter based on GNSS/INS integrated positioning when encountering the conditions of ambient occlusion,electromagnetic interference and sensor failure.Firstly,the improved Otsu method and the improved Hough transform algorithm were designed to extract the visual positioning information,and the relationship equation between the coordinate system of rice transplanter and camera was constructed to solve the position and pose values.Then,the federal Kalman filter algorithm with fault-tolerant function was used to fuse the visual positioning information with the positioning information output by GNSS and INS.Finally,tests were conducted in cement field and paddy field.Results showed that the accuracy of GNSS/INS/VNS combined positioning was similar to that of GNSS/INS integrated positioning in the open cement ground scene,but the average values of position error and heading error calculated by GNSS/INS/VNS combined positioning were 1.77 cm and 0.99°,respectively,which are 46.8%and 61.5%high-er than those by GNSS/INS integrated positioning method.In paddy field experiment,the average values of lateral deviation and heading deviation of navigation system decreased by 45.7%and 67.9%,respective-ly after visual supplement,with the average values of lateral deviation of 1.97 cm and heading deviation of 0.49°.It is indicated that the multi-sensor integrated positioning method based on visual supplement can ef-fectively reduce the positioning error and tracking deviation of navigation system,meeting the performance requirements of automatic driving operation of transplanter.