Research And Development Of Soybean Phenotype Data Platform System For Intelligent Breeding

Soybean is one of the important food crops.China is the largest soybean importer in the world,with an annual soybean import volume of more than 100 million tons.The huge gap between soybean production and demand,and the high dependence on imports have seriously threatened the national grain and oil security.It is imperative to accelerate the cultivation of high-yielding soybean varieties and improve the self-sufficiency rate of domestic soybeans.The study aims at the needs of soybean intelligent breeding,and constructs a soybean phenotype data platform system,which should not only reasonably store the massive phenotype data collected in the whole process of seed test,but also provide various services on demand based on the phenotype data,so as to provide large data support for breeding experts to design accurate breeding schemes.Based on the routine work flow of breeding experts,this study designs the platform system architecture and functional modules from top to bottom through detailed demand analysis,and builds the platform’s subsystems and integrates them into the platform system with a step-by-step refinement strategy.In order to store the soybean phenotype data collected by computer vision technology,based on the "Guidelines for the conduct of tests for distinctness,uniformity and stability—Soybean",the phenotypic properties of soybean at each growth stage recommended in the standard are expanded,and a phenotypic database model meeting the needs of seed test using artificial intelligence technology is designed.Each subsystem is only coupled at the data layer.The platform system has completed ten functional modules:material data management,breeding management,test management,variety knowledge base,phenotypic data management,data service,shared knowledge base,Target detection based on CV,registration and information maintenance,and system administrator module.The soybean variety knowledge base module is implanted with the ancestral and descendant tracing algorithm of soybean varieties based on map technology.Phenotypic data management is divided into field test phenotypes and indoor test phenotypes.Target detection based on CV module integrates the trained target detection model,plant phenotype extraction algorithm and strain recommendation algorithm.All functional modules are horizontally integrated and vertically connected to realize the informatization of the test process.Based on the effective organization,management and analysis of breeding data,provide accurate data for the subsequent expansion of more artificial intelligence services.The platform system is deployed in the cloud.The platform adopts B/S architecture to realize the decoupling and separation of the front and rear ends.The background module is built through Django,and the front-end design adopts react,ant design and other technologies.The front and rear ends are developed independently,reducing the coupling between the front and rear ends.At the same time,the system adopts modular design,and the coupling between modules only occurs at the data level,which ensures the high availability of the system.The platform system is oriented to soybean breeding experts and provides all-round,multi-dimensional and deep-seated functions such as shareable and easy to operate phenotypic data management,field planting planning,breeding list generation,variety pedigree tracing and strain intelligent recommendation.After more than one year of trial,the system has higher security,stability and expansibility,and speeds up the informatization process of domestic soybean excellent variety cultivation.