基于图像重建的黄连根系-土壤复合体力学特性仿真与试验  

Simulation and experiment on the mechanical properties of Coptis chinensis root-soil composites based on image reconstruction

作　　者：曾百功[1,2,3] 李敏 姚亮华 赵世龙 陈迅 王跃 刘凡一 谢守勇 ZENG Baigong;LI Min;YAO Lianghua;ZHAO Shilong;CHEN Xun;WANG Yue;LIU Fanyi;XIE Shouyong(College of Engineering and Technology,Southwest University,Chongqing 400715,China;Chongqing Innovation Engineering Center for Agricultural Technology,Chongqing 400715,China;Chongqing Key Laboratory of Agricultural Equipment in Hilly Area,Chongqing 400715,China)

机构地区：[1]西南大学工程技术学院,重庆400715 [2]重庆市农业技术创新方法工程技术中心,重庆400715 [3]丘陵山区农业装备重庆市重点实验室,重庆400715

出　　处：《农业工程学报》2023年第19期75-84,共10页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金项目(31601210,32201668);留学回国创业创新项目(cx2020118)。

摘　　要：针对黄连(Coptis chinensis)收获过程机械化程度低,根土复合体力学特性研究缺乏的问题,该研究以收获期黄连为研究对象,运用增量式运动恢复结构(incremental structure from motion,ISFM)和多视图立体(multi-view stereo,MVS)算法,通过多视角图像三维重构,将建立的高保真根系形态模型应用于不同条件下的根系-土壤复合体力学特性研究中。首先,通过田间采样和室内力学特性试验,得到根茎的平均剪切模量和泊松比,并测得不同深度的土壤和根土复合体的平均抗剪强度等物理力学参数。再以试验获得的入土深度为0~30 mm的土壤和根土复合体的平均抗剪强度作为响应值,采用Plackett-Burman试验和Central Composite试验等进行离散元仿真参数标定,并在EDEM软件中进行不同深度的土壤和根土复合体直剪试验数值模拟。结果表明:当土壤含水率在49.58%~62.96%、根系含水率在80%~230%的范围内时,在不同入土深度下,收获期黄连根土复合体离散元模型的内聚力和内摩擦角与试验结果的平均误差分别为3.48%和5.21%,标准差分别为3.44%和1.63%,验证表明通过图像重建根系模型可靠,并可用于模拟根土复合体力学行为。模拟方法可为黄连机械化收获关键工作部件设计和工作参数优化等提供理论依据。Chinese goldthread(Coptis chinensis)is one of the most important herbal medicines in eastern Asia.However,the manual harvesting cannot fully meet the large-scale production in recent years.Taking the Coptis chinensis at harvest period as the research object,this study aims to explore the mechanical properties of root-soil composites using image reconstruction.The incremental structure from motion(ISFM)was used to feature match the unordered images of Coptis chinensis roots,and then the geometric correction and triangulation were performed to restore the sparse point cloud structure.The relative pose was re-estimated using existing point clouds after local and global optimization.All camera poses and sparse 3D point clouds were output for the latter use.Then,the multi-view stereo(MVS)was used to estimate the depth map.Dense reconstruction of point clouds and dense 3D point clouds were obtained to contain the information,such as image color texture and a strong sense of reality.And then the images were converted into the polygonal data for optimization processing.A high-fidelity root morphology model was established using multi-view image reconstruction.The approach was applied into the mechanical research of root-soil composites under different conditions.The ring knife sampling was conducted on the rootless soil and in situ root-soil composites at different depths through field stratified sampling.The root profile was dug to measure the distribution range of the root system.It was found that the root system of Coptis chinensis was distributed mainly in the soil layer with a depth of 0-90 mm.The root quality first increased and then decreased,as the depth increased.The average shear modulus and Poisson's ratio of rhizomes were 4.08 MPa and 0.45,respectively,after tensile and shear mechanical testing.Direct shear tests were conducted on the soil and root-soil composites at depths of 0-30,>30-60,and>60-90 mm under loads of 50,100,150,and 200 kPa,respectively.The physical and mechanical parameters were obtained for t

关 键 词：图像重建 模型 仿真 根土复合体 直剪试验 黄连 

分 类 号：S233.75[农业科学—农业机械化工程]