摘要
花生是我国重要的油料作物,不同品种花生在生长特性、产量潜力和抗逆性等方面存在显著差异。网纹作为花生荚果的独特纹理特征,在形态、密度和分布上具有显著的品种特异性,是DUS测试的重要荚果性状,但现有研究对此利用不足。因此,本研究提出基于U-Net模型的花生网纹分割与多模态特征融合的品种识别框架。U-Net模型在对13个花生品种的网纹分割任务中表现优异,平均交并比为75.9%、准确率为89.2%,显著优于其他现有基础模型。进一步提取网纹图像的16个PCA降维特征,结合形态与颜色特征构建多模态数据集,采用SVM分类器实现品种识别,准确率达90.15%,较花生纹理、形态和颜色特征结合提升4.44%。研究首次证实花生网纹作为DUS测试性状的有效性,突破传统形态学的分析局限,为花生表型组学研究提供了可解释的方法,对推动精准育种和种质资源保护具有重要意义。
Peanut is an important oilseed crop in China,with significant differences among varieties in growth characteristics,yield potential,and stress resistance.The reticulation pattern on peanut pods,characterized by distinct varietal specificity in morphology,density,and distribution,serves as a key phenotypic indicator for DUS testing.However,existing studies have underutilized this trait.To address this,a U-Net based framework for peanut reticulation segmentation and multimodal feature fusion for variety identification was proposed.The U-Net model achieved outstanding performance in segmenting reticulation patterns through 13 peanut varieties,with a mean intersection over union of 75.9%and accuracy of 89.2%,significantly surpassing existing baseline models.Furthermore,16 PCA-reduced reticulation features were combined with morphological and color features to construct a multimodal dataset.Using the support vector machine classifier,the framework achieved a classification accuracy of 90.15%,representing 4.4%improvement over combinations of texture,morphology,and color features.This study is the first to confirm the validity of peanut reticulation as a DUS testing trait,overcoming limitations of traditional morphological analysis.The proposed method provides an interpretable approach for peanut phenomics research and holds significant value for advancing precision breeding and germplasm conservation.
作者
巩秀钇
踪姿艳
付华宇
张贺
纪翔
朱春雨
王聪
赵延伸
韩仲志
GONG Xiuyi;ZONG Ziyan;FU Huayu;ZHANG He;JI Xiang;ZHU Chunyu;WANG Cong;ZHAO Yanshen;HAN Zhongzhi(College of Science and Information Science,Qingdao Agricultural University,Qingdao 266109,China;College of Animation and Media,Qingdao Agricultural University,Qingdao 266109,China)
出处
《花生学报》
北大核心
2026年第1期23-33,共11页
Journal of Peanut Science
基金
山东省重点研发计划(2021LZGC026-05,2021TZXD003-003,2024LZGC006,2024TZXD037)
中央引导地方发展专项(23139-zyyd-nsh,22134-zyyd-nsh)
山东省科技型中小企业提升工程项目(2022TSGC1114,2021TSGC1016)
山东省泰山学者工程专项(2021-216)
农业农村部神农英才计划(202302186)。
