Acquisition And Analysis Image-based Phenotypic Traits Of Rice During The Whole Growth Period

In recent years,with the continuous updating and development of optical imaging systems and image processing technologies,the phenotype acquisition method for analyzing crop growth and development at the image level has been gradually recognized by researchers.As one of the most widely grown crops in the world,rice is not only the staple food but also the source of calorie intake for more than half of the world’s population,which occupies a very important position in China’s agricultural production.It is of vital importance that excavating the inner potential connection between genetic mechanisms and phenotypes with dynamic analysis of rice traits in high-throughput,non-destructive,and accurate based on high-throughput crop phenotyping facility in rice genetics and breeding research.In this study,the high-throughput visible light imaging platform is used to obtain rice images across the whole growth period,and the spatiotemporal multi-dimensional phenotypes obtained by a strategie for acquisition and analysis image-based spatio-temporal phenome in rice during the whole growth period proposed were applied to further analysis,and explore the association between phenotype and genotype,the main research work is as follows:(1)RGB image processing of potted rice.The traditional digital image processing method and the deep learning of image processing method were used in parallel to process the RGB images of the whole growth period of potted rice,and then the segmentation images of the whole rice plant,rice ear and rice stem were obtained respectively.(2)Extraction of spatiotemporal multi-dimensional phenotype of pot rice.Firstly,the whole growth period traits and stem-related traits of the whole rice plant and rice ear part were extracted,and the reliability of the traits was evaluated based on the dry weight traits of the whole rice plant,rice ear and stem obtained by manual measurement,respectively.The R~2 values were 0.88,0.73 and 0.67,respectively,which effectively confirmed the reliability and accuracy of phenotypes acquired.Then,the spatiotemporal multi-dimensional phenotype of potted rice was extracted by further analyzed on temporal dimension and the spatial dimension respectively.A total of 64 spatiotemporal phenotypes were obtained including the four of plant-related traits groups,panicle-related traits group,stem related traits groups and grain-related traits group in the spatial dimension and the three of phenological traits group,panicle development related traits and plant growth related traits in temporal dimension.(3)Yield modeling of spatiotemporal multi-dimensional phenotypes in pot rice.The method of stepwise linear regression was used to predict the final yield of pot rice with phenotypes in different spatial and temporal dimensions,and variables were be added into model or removed compared the preset criterion.Finally,it was found that the prediction effect of panicle-related traits traits and panicle development related traits were better than other trait sets,and the combination of the trait sets obtained by the preset criterion of model showed the best performance in predicting yield.(4)Analysis of differences in population structure of different genotypes and distribution trends of breeding region with spatiotemporal phenotypes of pot rice.Multiple tests were performed on samples of different groups based on the Kruskal-Wallis test and pairwise comparisons.The results of the distribution trend of the breeding region indicated that the traits TFN and Plant TPA of indica subspecies in breeding region at different latitude showed a negative correlation with latitude.The performance of the traits HS and Plant TPA of japonica subspecies in breeding region at different latitude were also shown the negative correlation of latitude and the breeding region.The results of analysis in difference of population structure indicated that the traits HS,Ave LW,Panicle YPA and Plant Ypar showed significant differences among different groups.(5)Genome-wide association analysis of spatiotemporal phenotypes of pot rice.Two QTLs on Chromosome 2 and Chromosome 4 were detected for the traits Plant Ypar and Stem Ypar respectively.55 QTLs and 49 QTLs respectively detected by Genome-wide association analysis with the principal components of the temporal traits and spatial traitsIn conclusion,a strategy for acquisition and analysis image-based spatio-temporal phenome in rice during the whole growth period were proposed in the study,and the feasibility of the strategy was verified through yield prediction,analysis of differences in population structure and distribution trends of breeding region with spatiotemporal phenotypes and genome-wide association analysis,which provides a new approach and a different thinking direction for the extraction and analysis of phenotypes of crops across the whole growth period.