FPheno2000:基于计算机视觉的食药用菌菌丝体表型采集与智能分析平台OA
FPheno2000:A Computer Vision-Based Platform for Phenotypic Data Acquisition and Intelligent Analysis of Edible and Medicinal Fungal Mycelia
为建立食药用菌菌丝表型高效、精准的采集与分析平台,设计并开发了基于计算机视觉的菌丝表型采集与智能分析平台FPheno2000.该平台具备高质量和规范化采集菌丝体图像的硬件设备,并结合条形码扫描系统实现数据的管理与追溯.菌丝表型数据智能化分析软件可搭配硬件使用,控制表型采集过程,并实现周长、面积、长边长、垂边长等7类菌丝表型的智能化提取与数据管理.在表型分析模块,构建了基于Yolov8和SAM的菌丝形态学表型与生长过程性状提取算法.以3株赤芝菌株为材料验证,通过使用菌丝图像进行测试,该SAM模型在零样本条件下对菌丝分割的平均交并比(mIOU)达到96.3%.提取的菌丝形态表型如面积、周长、垂边长、长边长等形态学表型与手动测量值高度相关,R2均在0.99以上.进一步基于时序数据构建Richard生长模型,构建了基于菌丝面积、周长、垂边长、长边长的菌丝生长模型曲线,拟合结果的R2均高于0.996,同时提取了菌丝最大生长速度、平均生长速度及生长持续期等多维度生长性状.FPheno2000平台的开发实现了菌丝体表型数据的采集和分析效率的提升,为食药用菌种质评价与精准育种提供了可靠工具.
To facilitate high-throughput and precise acquisition and analysis of mycelial phenotypes in edible and medicinal fungi,we have developed FPheno2000,an advanced computer vision-based platform for mycelial phenotyping and intelligent data analysis.This platform combines specialized hardware designed for high-quality,standardized imaging of mycelia with an integrated barcode scanning system to ensure efficient data management and comprehensive traceability.The associated intelligent phenotyping software interfaces with the hardware to automate image acquisition and to extract and manage seven distinct categories of mycelial phenotypic traits,including perimeter,area,long-axis length,and short-axis length.For the phenotypic analysis component,we implemented algorithms for extracting mycelial morphological features and growth-process traits,leveraging the YOLOv8 framework alongside the Segment Anything Model(SAM).Validation experiments conducted on three Ganoderma sichuanense strains demonstrated that the SAM-based model achieved a mean intersection-over-union(mIoU)of 96.3%for mycelial segmentation under zero-shot conditions.The morphological traits extracted such as area,perimeter,short-axis length,and long-axis length-exhibited strong concordance with manual measurements,with coefficients of determination(R2)exceeding 0.99 across all traits.Additionally,time-series data were modeled using the Richards growth function to generate growth curves based on area,perimeter,and axis lengths,yielding R2 values greater than 0.996.From these growth curves,key growth parameters including maximum growth rate,average growth rate,and growth duration were derived.In summary,FPheno2000 significantly enhances the efficiency and accuracy of mycelial phenotype data acquisition and analysis,offering a robust tool for fungal germplasm evaluation and precision breeding initiatives.
李卓识;刘云飞;刘唯真;付永平;莫智康;皮嘉龙;袁晓辉
吉林农业大学植物保护学院、菌物学院,长春 130118||吉林农业大学信息技术学院,长春 130118吉林农业大学信息技术学院,长春 130118武汉理工大学计算机与人工智能学院,武汉 430070吉林农业大学植物保护学院、菌物学院,长春 130118武汉理工大学计算机与人工智能学院,武汉 430070武汉理工大学计算机与人工智能学院,武汉 430070吉林省菌物表型组学重点实验室,长春 130118
农业科技
食药用菌表型组学分析平台大模型
edible and medicinal fungiphenomicsanalysis platformlarge-scale model
《菌物研究》 2026 (4)
328-339,12
国家重点研发计划项目(2023YFD1201600),吉林省科技发展计划项目(20240305025YY)
评论