灵芝子实体三维表型参数提取方法  

A method for extracting three-dimensional phenotypic parameters of fruiting bodies of Ganoderma lingzhi

作　　者：茹晨雷 熊晖 潘在晨 杨继鸿 徐靖 李振皓 RU Chenlei;XIONG Hui;PAN Zaichen;YANG Jihong;XU Jing;LI Zhenhao(Zhejiang Shouxiangu Botanical Drug Institute Co.,Ltd.,Hangzhou 321200,Zhejiang,China;Pharmaceutical Informatics Institute,College of Pharmaceutical Sciences,Zhejiang University,Hangzhou 310058,Zhejiang,China;Zhejiang Key Laboratory of Biological Breeding and Exploitation of Edible and Medicinal Mushrooms,Jinhua 321200,Zhejiang,China;Zhejiang Shouxiangu Pharmaceutical Co.,Ltd.,Jinhua 321000,Zhejiang,China)

机构地区：[1]浙江寿仙谷植物药研究院有限公司,浙江杭州321200 [2]浙江大学药学院药物信息学研究所,浙江杭州310058 [3]全省食药用菌生物育种与综合开发利用重点实验室,浙江金华321200 [4]浙江寿仙谷医药股份有限公司,浙江金华321000

出　　处：《菌物学报》2025年第2期149-159,共11页Mycosystema

基　　金：浙江省农业(食药用菌)新品种选育重大科技专项(2021C02073);浙江省重点农业企业研究院(2017Y20001)。

摘　　要：三维表型结构信息在灵芝生长过程监测和精准育种中发挥着重要作用。为快速、无损、精准地获取灵芝表型信息,本研究提出了一种基于多视角图像三维重构的灵芝子实体三维表型参数提取方法。使用智能手机获取待测子实体的网格模型;将网格模型转化为点云模型,对其进行尺寸校正;沿Z轴进行点云切片,根据切片位置提取子实体点云,使用基于颜色的区域分割算法提取粗糙孔口表面部件;使用DBSCAN(density-based spatial clustering of applications with noise)算法和最近邻分配算法分割精细菌盖、孔口表面和菌柄部件;最后从不同部件中提取菌盖厚度、菌盖横轴、菌柄长度、子实体体积、菌盖颜色、孔口表面颜色、菌盖投影面积等表型参数。结果表明,与真实值相比,使用本研究方法计算的菌盖厚度、菌盖横轴、菌柄长度和子实体体积的均方根误差(root mean squared error,RMSE)分别为1.10 mm、4.55 mm、1.53 mm和0.038 L,决定系数(R^(2))分别为0.995、0.972、0.986和0.981。所提方法可以精准提取灵芝子实体三维表型结构信息,并有望拓展应用至其他食药用菌子实体。Three-dimensional phenotypic structural information plays an important role in monitoring the growth process and precise breeding of Ganoderma lingzhi.In order to rapidly,non-destructively,and accurately obtain phenotypic information,a method was proposed for extracting three-dimensional phenotypic parameters of fruiting bodies of G.lingzhi based on multi-view three-dimensional reconstruction.A mesh model of the fruiting body was obtained by using smartphone.The mesh model was converted into a point cloud model and underwent size calibration.Point cloud slicing along the Z-axis was performed to extract the parts of fruiting body and color-based region segmentation algorithm was used to extract rough pore surface parts.The DBSCAN(density-based spatial clustering of applications with noise)algorithm and nearest neighbor assignment algorithm were used to segment fine pileus,pore surface,and stipe components.Finally,phenotypic parameters of pileus thickness,pileus transverse axis,stipe length,fruiting body volume,pileus color,pore surface color,pileus projection area,etc.were extracted from different components.Results showed that compared with the true values,the root mean squared errors(RMSEs)of pileus thickness,pileus transverse axis,stipe length,and fruiting body volume calculated by using this method were 1.10 mm,4.55 mm,1.53 mm,and 0.038 L,respectively,and the determination coefficients(R^(2))were 0.995,0.972,0.986,and 0.981,respectively.The proposed method can accurately extract three-dimensional phenotypic structural information of fruiting bodies of G.lingzhi,and is expected to be applied to other edible and medicinal fungal fruiting bodies.

关 键 词：灵芝 基于多视角图像三维重构 部件分割 三维表型参数 

分 类 号：S567.31[农业科学—中草药栽培]