摘要
针对现有果园导航技术存在GNSS信号遮挡、视觉受强光干扰及3D激光雷达成本高昂等问题,进行基于单线激光雷达果园导航路径提取方法的研究。通过动态调整阈值剔除杂草干扰点云,并聚类果树点云,结合单协方差自适应kalman滤波算法与最小二乘法平滑果树定位点后拟合导航路径。仿真结果表明:改进算法在0.5~1.5 m·s^(-1)5种速度下路径拟合平均准确率为96.5%,平均最大偏差与平均偏差分别为0.068 m、0.035 m,相较传统算法分别降低4.1%、10.7%,方差降低55.6%~62.7%;真实行间环境中,在3种起点偏移(0 m,±0.15 m)下路径拟合平均准确率为89%,平均最大偏差、平均偏差分别为0.198 m、0.108 m,相较传统算法分别降低12.3%、21.6%,方差降低19.3%~32.6%,验证改进算法在复杂的行间环境中提取导航路径的实用性。
To address the challenges of GNSS signal occlusion,visual interference under strong light,and high costs of 3D LiDAR in existing orchard navigation technologies,this study proposes a navigation path extraction method based on single-line LiDAR.By dynamically adjusting thresholds to remove weed-interfered point clouds and cluster tree trunk point clouds,combined with a single-covariance adaptive Kalman filtering algorithm and least squares method,the positioning points are smoothed and navigation paths are fitted.Simulation results demonstrate that the improved algorithm achieves an average path fitting accuracy of 96.5%under five speeds(0.5~1.5 m·s^(-1)),with average maximum and mean deviations of 0.068 m and 0.035 m,respectively,representing reductions of 4.1%and 10.7%compared to conventional algorithms,while variances decrease by 55.6%~62.7%.In real inter-row environments,under three initial position offsets(0 m,±0.15 m),the average accuracy is 89%,with average maximum and mean deviations of 0.198 m and 0.108 m,respectively,corresponding to reductions of 12.3%and 21.6%in deviations and 19.3%~32.6%in variances.These results validate the practicality of the improved algorithm for extracting navigation paths in complex orchard environments.
作者
史志康
黄鑫
钱金源
易克传
SHI Zhikang;HUANG Xin;QIAN Jinyuan;YI Kechuan(College of Intelligent Manufacturing,Anhui Science and Technology University,Chuzhou 239000,China)
出处
《黑龙江工程学院学报》
2026年第1期26-34,共9页
Journal of Heilongjiang Institute of Technology
基金
安徽省教育厅高校科研项目(2024AH050305)
安徽科技学院引进人才项目(JXYJ202204)。
关键词
果园导航
单线激光雷达
动态阈值
自适应卡尔曼滤波
路径提取
orchard navigation
single-line LiDAR
dynamic threshold
adaptive Kalman filtering
path extraction
