基于实例分割的休眠期枣树修剪枝参数提取方法  

Extraction method of pruning parameters of dormant jujube tree based on instance segmentation

作　　者：马保建 陈棒棒 李学志 蒋焕煜 Ma Baojian;Chen Bangbang;Li Xuezhi;Jiang Huanyu(College of Mechanical and Electrical Engineering,Xinjiang Institute of Technology,Aksu,843100,China;College of Biosystems Engineering and Food Science,Zhejiang University,Hangzhou,310058,China)

机构地区：[1]新疆理工学院机电工程学院,新疆阿克苏843100 [2]浙江省农业智能装备与机器人重点实验室,杭州市310058

出　　处：《中国农机化学报》2024年第10期54-59,68,共7页Journal of Chinese Agricultural Mechanization

基　　金：新疆维吾尔自治区自然科学基金(2022D01C357)。

摘　　要：为实现休眠期枣树选择性修剪,针对自动化剪枝过程中枣树枝干识别与参数提取困难的问题,提出一种基于实例分割神经网络的枝干识别与参数自动提取的方法。首先,通过前期搭建的视觉系统获取2个角度下的点云,并基于骨架点重建完整的枣树点云,利用CloudCompare V 2.11软件对枣树三维点云进行手工标注,构建带语义信息的枣树点云数据集,将标注完的376棵枣树,按照8∶2的比例分为训练集和验证集;其次,结合JSNet实例分割网络对枣树点云进行分割,同时对比分析不同自然环境对实例分割精度的影响;最后,提出枣树修剪枝直径和长度参数自动提取的方法。试验结果表明:休眠期枣树主干的分割精度为96%,而修剪枝的分割精度为77%,并且不同自然环境对分割精度影响较小。枣树修剪枝的直径拟合值与实际测量值误差范围在2 mm以内,且其长度拟合值与实际值的误差范围在1 cm以内,为后续准确确定枣树剪枝点的位置提供数据依据。In order to achieve selective pruning of dormant jujube tree,the identification and parameter extraction method of pruning branches of dormant jujube tree was proposed based on instance segmentation neural network,in view of the difficulty of extracting pruning parameters in automatic pruning process.Firstly,point cloud from two angles were obtained through the vision system built earlier,and a complete point cloud of date tree was reconstructed based on skeleton points.The three‑dimensional point cloud of date tree was manually labeled by CloudCompare V 2.11 software,and dataset of date tree with semantic information was constructed.The labeled 376 trees were divided into training and validation dataset according to the ratio of 8∶2,and then instance segmentation of dormant jujube tree was carried out by JSNet neural network.The influence of different natural environment for the segmentation accuracy was analyzed.Secondly,the fitting calculation method for diameter and length parameters of pruning branches was proposed.The experimental results showed that the segmentation accuracy of jujube tree trunk was 96%,and that of pruning branches was 77%,which was less affected by the natural environment.The relative error between the diameter fitting result and the actual value was less than 2 mm,and the error between the length value and actual measurement value was less than 1 cm,which provided a basis for the accurate determination of the location of pruning point of jujube tree.

关 键 词：休眠期枣树 实例分割 深度学习 三维点云 修剪枝 

分 类 号：S225.93[农业科学—农业机械化工程] TP391[农业科学—农业工程]