Two-dimensional Semantic Morphological Feature Extraction and Atlas Construction of Maize Ear Leaves

Hongli Song ¹ Weiliang Wen ² Ying Zhang ² Yanxin Zhao ² Xinyu Guo ^2* Chunjiang Zhao ²

• ¹ Northwest A&F University, Xianyang, Shaanxi Province, China
• ² Beijing Academy of Agricultural and Forestry Sciences, Beijing, China

Maize ear leaves have important roles in photosynthesis, nutrient partitioning and hormone regulation.The morphological and structural variations observed in maize ear leaves are numerous and contribute significantly to the yield. Nevertheless, research on the fine-scale morphology of maize leaves is less, particularly the quantitative methods to characterize the morphology of leaves in two-dimensional (2D) space is absent. This makes it challenging to accurately identify 2D leaf shape of their cultivars. Therefore, this study presents the methods of 2D semantic morphological feature extraction and atlas construction, with the ear leaf in silking stage of maize association analysis population serving as an example. A three-dimensional (3D) digitizer was employed to obtain data from 1,431 leaves belonging to 518 inbred lines. The data was then processed using mesh subdivision and planar parameterization to create 2D leaf models with areapreserving characteristics. Additionally, averaged 2D leaf models of all the inbred lines were constructed, and 29 2D leaf features were quantified. Based on this, 11 features were extracted as semantic features of 2D leaf shape through clustering and correlation analysis. A comprehensive 2D leaf shape indicator 𝐿 based on the 11 semantic features was proposed, and a 2D leaf shape atlas was constructed in accordance with the 𝐿 ordering. Inbred line identification of 2D leaf shape in maize was achieved using the atlas. The results of maize leaf inbred line identification can determine the probability that the corresponding true inbred line ranked within the top 10 of the predicted results is 0.706, within the top 20 is 0.810, and within the top 45 is 0.900. This enables the generation of the corresponding maize 2D leaf shape through the matching of semantic features. The methodology presented in this study offers novel insights into the construction of semantic models for the morphology of maize and the identification of cultivars. It also provides a theoretical and technical foundation for the generation and drawing the leaf shape based on semantic 2D morphological and structural features.