摘要
为提高金银花采摘机器人在自然环境下的识别与采摘精度,针对金银花花簇生长分散的特性,提出Top—Bottom—Layer—Points算法。该算法精准判定花簇最顶层与最底层点集,并有效计算采摘压下量。通过深度相机捕获RGB—D图像,结合深度学习算法实现花朵轮廓高精度分割,以此构建花簇三维点云。在此基础上,根据最顶层点集拟合的平面法向量确定最佳采摘方向,同时利用最底层点集与模拟退火算法精确锁定最优采摘点,模拟人工采摘动作以最大程度减少对金银花的损伤。试验结果显示,该方法在自然环境下对金银花的识别准确率高达93%,采摘成功率达93.4%,且损坏率低于6%。不仅满足农业领域对金银花高效、有效、精准采摘的迫切需求,更为金银花智能化采摘技术的发展提供关键的技术支撑。
In order to improve the recognition and positioning picking accuracy of the honeysuckle picking robot in natural the environment,the Top—Bottom—Layer—Points algorithm was proposed based on the dispersed growth characteristic of honeysuckle.This algorithm accurately determines the top-most points and the bottom-most points of honeysuckle and calculates the picking pressure reduction.The image information of a honeysuckle was collected by a depth camera,and the high-precision segmentation of honeysuckle contours was achieved and combined with deeplearning algorithms to construct the 3D point clouds of a honeysuckle cluster.To simulate manual picking action to minimize damage to honeysuckle,the normal vector of the best-fit plane for the topmost points was taken as the optimal picking direction.The optimal picking point was determined using the bottom-most points combined with a simulated annealing algorithm.The experimental results show that in natural environments,the recognition accuracy of honeysuckle is 93%at the lowest,and the picking success rate reaches 93.4%,with the damage rate less than 6%.This study not only meets the urgent demand for efficient,effective,and high-precision picking of honeysuckle in agriculture,but also provide powerful technical support for the realization of intelligent picking of honeysuckle.
作者
秦亚敏
于浩
贾英新
张震
程豪
李永伟
Qin Yamin;Yu Hao;Jia Yingxin;Zhang Zhen;Cheng Hao;Li Yongwei(Hebei Mechatronics Intermediate Pilot Production Base Co.,Ltd.,Shijiazhuang,050081,China;Hebei Intelligent Measurement and Control Technology Innovation Center,Shijiazhuang,050081,China;Zhejiang Saile New Energy Materials Co.,Ltd.,Jinhua,321000,China)
出处
《中国农机化学报》
北大核心
2026年第1期160-166,240,共8页
Journal of Chinese Agricultural Mechanization
基金
河北省科学院科技计划项目(24A11,22715)
石家庄市高价值专利组合培育项目(202401)。
关键词
金银花
精确定位
三维点云
采摘方向
采摘点
honeysuckle
accurate location
3D point clouds
picking direction
picking point
