Low Field-NMR-Based Study On The Detection Method Of Ginkgo Seed Germination Quality

With the development of Low Field Nuclear Magnetic Resonance(LF-NMR)and Magnetic Resonance Imaging(MRI)technology,as well as the localization of LF-NMR equipment,LF-NMR has been widely applied in the fields of agriculture and forestry.In this study,machine vision,deep learning,and image processing technologies were employed to grade the activity levels and detect embryo defects of ginkgo seeds at different levels of viability.Based on NMR and MRI,the germination quality of ginkgo seeds was investigated.The results showed that LF-NMR and MRI technology can accurately detect the activity level of ginkgo seeds and identify the presence of embryo defects,providing a new approach for the quality control of ginkgo seeds.This study provides a valuable reference for the practical application of LF-NMR and MRI technology in the agricultural industry.This study investigates the relationship between ginkgo seed viability and moisture content by using the drying method to determine the moisture content of ginkgo seeds with different levels of viability.Additionally,artificial aging was employed to prepare batches of ginkgo seeds with varying activity levels for germination tests,measuring indicators such as germination vigor,germination rate,and germination index during the growth process.Based on the results,the relationship between ginkgo seed viability and moisture content was elucidated.This study provides a valuable reference for the practical application of ginkgo seeds in the fields of agriculture and forestry.This study investigates the changes in water absorption rate of ginkgo seeds with different levels of viability using LF-NMR technology.The partial least squares method was used to establish a prediction model for ginkgo seed moisture content based on LF-NMR data,which revealed the relationship between nuclear magnetic signal and moisture content.The water absorption rate of ginkgo seeds with different levels of viability during germination was then determined.The results showed that the water absorption rate of ginkgo seeds gradually decreased as the aging days increased and the viability decreased during the germination process.This study provides valuable insights into the relationship between ginkgo seed viability and water absorption rate,which can aid in the practical application of ginkgo seeds in the agricultural and forestry industries.Research on determination of ginkgo biloba seed vigor level based on machine learning.After determining the T2 inversion spectrum characteristic values related to vigor indicators,multiple linear and nonlinear BP neural network regression prediction models were established.Comparative studies showed that the fitting ability of the BP neural network regression model was superior to that of the linear model,with a lower misjudgment rate and an overall accuracy of 87.78%.The vigor of the seeds was evaluated using germination potential and germination index.Comparative studies showed that the BP neural network regression model based on germination index as an evaluation indicator could be used for the determination of ginkgo biloba seed vigor.Research on Detection Method of ginkgo Seed Embryo Defects Based on Image Processing and Deep Learning.The seeds of ginkgo were divided into three categories according to whether there was embryo inside and whether the embryo was rotten: intact with embryo,intact without embryo,and embryo rot.MRI images of ginkgo seeds were collected and preprocessed.Combined with transfer learning,five deep learning models including Alex Net,Res Net-18,Res Net-50,Res Ne Xt-50,and Res Ne Xt-101 were used to classify and detect the quality of ginkgo seed embryo.The results showed that all five models could classify the quality of ginkgo seeds,and the Res Ne Xt series models had better overall performance than the other three models.Among them,the Res Ne Xt-101 model had the best performance,with precision rates of 98.8633%,97.2145%,and 99.5347% for identifying the three types,respectively,and the highest overall model accuracy of 98.3086%.Compared with the Res Ne Xt-101 model,although the Res Net-50 model had slightly lower accuracy,it had shorter training time and was a more ideal model for classifying the quality of ginkgo seed embryo.This study demonstrates that LF-NMR technology can be used for non-destructive detection of germination quality in ginkgo biloba seeds,providing a new method to improve seeding quality and increase field emergence rates,as well as offering theoretical basis and technical support for non-destructive testing of seeds with woody seed coats.This has important implications for improving seed sowing methods and enhancing overall seedling production.