一种果园特征识别与定位的导航方法
- 分享到:
摘要：
针对高山丘陵等复杂果园环境中,植保无人机导航作业精度低,果实特征位置估计误差大,且果树特征提取中噪声多等问题,提出了一种基于四元数的平方根容积卡尔曼滤波算法,该算法以姿态四元数误差及陀螺仪漂移误差为状态量,以 SINS/SLAM组合导航的姿态四元数为量测量,并进行时间更新与量测更新;同时,采用 Kmeans 等方法处理田园环境信息,实现了较高精度的果树特征的提取,基于四元数的平方根 CKF算法,既解决了传统四元数的规范化问题,降低了传统四元数的平方根 UKF算法的状态维数及计算复杂度,与四元数 SRUKF.四元数 SRCDKF算法比较,仿真实验结果表明新算法估计横滚角、俯仰角、航偏角误差均值分别为 0.05°0.08°0.03°,误差均为最小,且算法精度较四元数 SRUKF-SLAM算法提高了 30%左右,在较大初始角误差条件下,进行对 SRCKE.CKE.SRUKE3种滤波算法的估计误差对比,实验表明 SRCKE算法均具有最高的滤波精度,目滤波收敛速度最快、稳定性最好.
Aiming at the problems such as low navigation accuracy of plant protection UAV, large estimation error of truit feature position, and excessive nolse in trut tree feature extraction in complex orchard environments such as mountains and hills, a quaternion based square root cubature Kalman filtering algorithm was proposed, which takes attitude quatermion error and gyroscope drit error as state variables. The attitude quaternion of SINS/SlAM integrated navigation is measured, and the time and measurement update are carried out, At the same time, Kmeans and other methods are used to process the pastoral environment information, and the characteristics of fruit trees are extracted with high precision, The square root CKf algorithm based on quaternions not only solves the normalization problem of traditional quaternions, but also reduces the state dimension and computational complexity of the square root UKF algorithm of traditional quaternions. Compared with the quaternion SRUKF and SRCDKF algorithms, simulation results show that the new algorithm estimates the average errors of roll angle, pitch angle, and yaw angle, which are 0.05°, 0.08°, and 0.03°, respectively, with the smallest errors. Moreover, the algorithm accuracy is improved by about 30% compared with the quaternion SRUKF-SLAM algorithm. Under the condition of large initial angular error, the estimation errors of three filtering algorithms, SRCKF, CKF, and SRUKF, were compared. The experiment showed that SRCKF algorithm has the highest filtering accuracy, the filtering convergence speed is the fastest and the stability is the best.
作者：
祝朝坤，杜雪，谭开拓，伍龙
Wang Dandan，Zhu Zhaokun，Du Xue，Tan Kaituo，Wu Long
机构地区：
哈尔滨工程大学智能科学与工程学院；淮南师范学院机械与电气工程学院；郑州工商学院信息工程学院
引用本文：
王丹丹，祝朝坤，杜雪等。一种果园特征识别与定位的导航方法[J].河南师范大学学报(自然科学版),2025.53(5):113-120.( Wang Dandan,Zhu Zhaokun, Du Xue, et al.A navigation method for orchard feature recognition and localization[J],Journal of Henan Normal University(Natural Science Edition),2025,53(5):113-120.DOI：10.16366/j.cnki.1000-2367.2024.08.10.0001.)
基金：
安徽省优秀中青年教师培育项目；安徽省质量工程项目
关键词：
四元数；状态模型；平方根容积卡尔曼滤波；姿态估计；数值稳定性数
squaternion; state model; square root cubature Kalman filter; attitude estimation; numerical stability
分类号：
U666.1
