基于图像识别技术的金针菇表型高通量采集与分析

High-throughput phenotyping collection and analysis of Flammulina filiformis based on image recognition technology

金针菇Flammulina filiformis作为国内产销量大、工厂化生产程度高的食用菌种类之一,其育种工作因表型信息采集效率低下而受到限制。本研究以金针菇子实体可见光图像为信息来源,基于育种工作对表型信息的需求,利用图像识别技术和深度学习模型,提出了金针菇高通量表型信息采集分析方法,并开发了相应分析系统软件。利用该套方法结合金针菇子实体的可见光扫描图像,可以自动化大批量计算金针菇子实体的菌盖形状、面积、着生位置、颜色和菌柄长度、宽度、颜色等表型信息。本研究提出的金针菇表型采集与分析方法操作便捷,能大幅减少观测误差,提高表型信息可靠性,提升育种工作效率,可一定程度上应用于其他外形相近的食用菌或作物育种工作;同时获得的数据有助于金针菇表型信息的定量化分析和标准化表型数据集的建立。

Flammulina filiformis is a kind of edible fungus species that has been largely factory-produced domestically.However,its breeding work is still limited due to the low efficiency of phenotyping collection.In this study,the visible images of F.filiformis fruiting bodies were used as informative source,and a method for phenotyping collection and analysis was proposed by use of image recognition techniques and deep learning models to meet the needs of phenotypic information in breeding.A software for this method was consequently developed.Based on the visible light scanning images of F.filiformis fruiting bodies,this method is able to automatically calculate the phenotypic information of fruiting bodies in large quantities,including cap shape,area,position and color,and stipe length,width and color.The proposed method can greatly reduce observation errors,increase the reliability of phenotypic information,and extend to breeding of other edible fungi with similar appearance.The data obtained are instructive to quantitative analysis of phenotypic information of F.filiformis and useful to establish standardized phenotype data sets.