基于改进U-Net++模型的油菜pol TCMS温敏两系育性等级鉴定及温度育性关系的量化研究

doi:10.3724/SP.J.1006.2025.44198

摘要/Abstract

摘要：

油菜(Brassica napus L.)是全球重要的油料作物，其杂种优势的利用在提升产量与环境适应能力方面发挥了关键作用。温敏型波里马细胞质雄性不育系(pol TCMS)因其育性受温度影响，具备一系两用的独特优势，已成功应用于两系育种。雄蕊、雌蕊长度是生产实践中进行育性等级划分的主要依据，然而育性等级的主观划分方法易受人为因素干扰，最终影响表型考察及精细定位的结果。为此，本研究提出了一种育性等级鉴定的新方法，即基于改进U-Net++深度学习模型的图像语义分割法。以pol TCMS温敏两系及pol CMS稳定不育系构建F₂分离群体，首先，获取分离群体中不同育性等级花器官的图像，标注构建数据集；其次，选取U-Net++图像语义分割方法，通过优化编码器和解码器结构并引入通道注意力模块，提升模型的分割精度；最后，以不同育性等级的图像进行训练和测试。结果表明，改进后的模型在不同育性等级花器官的分割任务中，平均交并比为92.02%，精确率为98.94%，召回率为98.84%，F₁分数为98.87%，优于其他分割模型方法，该模型能很好识别出不同育性等级的花器官。基于分割结果获得原位实际长度，与人工测量的长度相比，预测值与实测值的决定系数R²为0.989，均方根误差(RMSE)为0.142 mm，Spearman相关系数为0.983，可以实现不同育性等级表型参数的准确测量。此外，通过分析量化温度与育性(雄蕊/雌蕊比值)的关系发现，育性随单花开花前9 d的温度变化而波动。本研究验证了温度对油菜育性的关键影响，为深入解析油菜温敏特性及相关基因定位研究提供了新方法和技术支持。

关键词: 油菜, Pol TCMS, U-Net++分割模型, 温度, 育性等级, 基因定位

Abstract:

Rapeseed (Brassica napus L.) is a globally important oil crop, and the exploitation of its hybrid vigor has played a crucial role in enhancing yield and environmental adaptability. The thermosensitive Polima cytoplasmic male sterile (pol TCMS) line has a unique dual-purpose advantage, as its fertility is temperature-dependent, making it highly valuable for two-line breeding systems. In agricultural practice, the classification of fertility status is primarily based on the relative lengths of stamens and pistils. However, traditional fertility grading methods rely on subjective assessment, which is prone to human error, ultimately affecting phenotypic investigations and precise trait mapping. To address this issue, we propose a novel approach for fertility classification based on image semantic segmentation using an improved U-Net++ deep learning model. An F₂ segregating population was developed using the pol TCMS thermosensitive line and the pol CMS stable sterile line. First, floral organ images representing different fertility levels within the segregating population were collected and annotated to construct a dataset. Next, we optimized the U-Net++ model by refining its encoder-decoder architecture and integrating a channel attention module to enhance segmentation accuracy. Finally, the model was trained and tested on images corresponding to different fertility levels. Experimental results demonstrated that the improved model achieved a mean intersection-over-union (mIoU) of 92.02%, precision of 98.94%, recall of 98.84%, and an F₁ score of 98.87%, outperforming other segmentation models. The model effectively distinguished floral organs across different fertility grades and enabled in situ measurements of organ length based on segmentation results. Compared with manual measurements, the predicted values exhibited a coefficient of determination (R²) of 0.989, a root mean square error (RMSE) of 0.142 mm, and a Spearman correlation coefficient of 0.983, demonstrating high accuracy in phenotypic parameter estimation across different fertility levels. Furthermore, we quantitatively analyzed the relationship between temperature and fertility (stamen/pistil ratio), revealing that fertility fluctuates in response to temperature variations during the nine days preceding anthesis in individual flowers. This study confirms the key role of temperature in regulating fertility and provides a novel methodological framework and technical support for in-depth analyses of temperature-sensitive traits and gene mapping in rapeseed.

Key words: rapeseed, Pol TCMS, U-Net++ segmentation model, temperature, fertility grade, gene location

李世鹏, 陈才武, 张晶, 吕恬, 傅廷栋, 易斌. 基于改进U-Net++模型的油菜pol TCMS温敏两系育性等级鉴定及温度育性关系的量化研究[J]. 作物学报, 0, (): 0-.

LI Shi-Peng, CHEN Cai-Wu, ZHANG Jing, LYU Tian, FU Ting-Dong, YI Bin. Identification of fertility levels and quantification of the temperature-fertility relationship in rapeseed pol TCMS lines using an improved U-Net++ model[J]. Acta Agronomica Sinica, 0, (): 0-.

参考文献

[1] 刘杰, 黄学辉. 作物杂种优势研究现状与展望. 中国科学: 生命科学, 2021, 51: 1396–1404.

Liu J, Huang X H. Advances and perspectives in crop heterosis. Sci Sin Vitae, 2021, 51: 1396–1404 (in Chinese with English abstract).

[2] 袁国强, 陶军, 何员江, 罗江陶, 蒲宗君, 李生荣, 任勇. 杂交小麦研究进展与展望. 四川农业大学学报, 2023, 41: 973–980.

Yuan G Q, Tao J, He Y J, Luo J T, Pu Z J, Li S R, Ren Y. Research progress and prospect on hybrid wheat. J Sichuan Agric Univ, 2023, 41: 973–980 (in Chinese with English abstract).

[3] 汪邑晨, 朱本顺, 周磊, 朱骏, 杨仲南. 光/温敏核不育系的不育机理及两系杂交稻的发展与展望. 中国水稻科学, 2024, 38: 463–474.

Wang Y C, Zhu B S, Zhou L, Zhu J, Yang Z N. Sterility mechanism of photoperiod/thermo-sensitive genic male sterile lines and development and prospects of two-line hybrid rice. Chin J Rice Sci, 2024, 38: 463–474 (in Chinese with English abstract).

[4] Xiao Q, Wang H D, Chen H, Chen X H, Wen J, Dai C, Ma C Z, Tu J X, Shen J X, Fu T D, et al. Molecular analysis uncovers the mechanism of fertility restoration in temperature-sensitive polima cytoplasmic male-sterile Brassica napus. Int J Mol Sci, 2021, 22: 12450.

[5] 杨光圣. 甘蓝型油菜雄性不育的分类和综合利用研究. 华中农业大学博士学位论文, 湖北武汉, 1998.

Yang G S. Study on Classification and Comprehensive Utilization of Male Sterility in Brassica napus L. PhD Dissertation of Huazhong Agricultural University, Wuhan, Hubei, China, 1998 (in Chinese with English abstract).

[6] 袁美. 甘蓝型油菜生态型细胞质雄性不育两用系温度调控机理的初步研究. 华中农业大学博士学位论文, 湖北武汉, 2002.

Yuan M. Preliminary Studies on Temperature Regulation on the Ecotypical Male Sterile-fertile Line of Brassica napus L. PhD Dissertation of

Huazhong Agricultural University, Wuhan, Hubei, China, 2002 (in Chinese with English abstract).

[7] 杨立勇. Pol TCMS小孢子培养体系的建立与温敏基因的QTL定位. 华中农业大学博士学位论文, 湖北武汉, 2005.

Yang L Y. Establishment of Microspore Culture System for Pol TCMS and QTL Mapping of Its Temperature-sensitive Genes. PhD Dissertation of

Huazhong Agricultural University, Wuhan, Hubei, China, 2005 (in Chinese with English abstract).

[8] 肖清. 甘蓝型油菜高密度SNP育种芯片的开发及其在Pol-CMS温敏不育研究中的应用. 华中农业大学博士学位论文, 湖北武汉, 2022.

Xiao Q. Development of High-density SNP Breeding Chip and Its Application in Pol-CMS Thermosensitive Sterility Research in Brassica napus.

PhD Dissertation of Huazhong Agricultural University, Wuhan, Hubei, China, 2022 (in Chinese with English abstract).

[9] 张小强, 熊博莅, 匡纲要. 一种基于变化检测技术的SAR图像舰船目标鉴别方法. 电子与信息学报, 2015, 37(1): 63–70.

Zhang X Q, Xiong B L, Kuang G Y. A ship target discrimination method based on change detection in SAR imagery. J Electr Inf Technol, 2015,

37(1): 63–70 (in Chinese with English abstract).

[10] 徐志祥, 孙文博, 邢立东, 高东, 赵炎. 基于改进Canny边缘检测的堆叠货箱分割定位方法. 测控技术, 2024, 43(3): 28–33.

Xu Z X, Sun W B, Xing L D, Gao D, Zhao Y. Segmentation and location method of stacked containers based on improved canny edge detection.

Meas Contr Technol, 2024, 43(3): 28–33 (in Chinese with English abstract).

[11] Zeng Y Z, Zhao Y Q, Liao S H, Liao M, Chen Y, Liu X Y. Liver vessel segmentation based on centerline constraint and intensity model. Biomed Signal Process Contr, 2018, 45: 192–201.

[12] Fukushima K. Neocognitron: a self organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biol Cybern, 1980, 36: 193–202.

[13] Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for biomedical image segmentation. In: Ronneberger O, Fischer P, Brox T, eds. Medical Image Computing and Computer-assisted Intervention–MICCAI 2015. Cham: Springer International Publishing, 2015. pp 234–241.

[14] Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation. In: Long J, Shelhamer E, Darrell T, eds. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Computer Vision Foundation, 2015. pp 3431–3440.

[15] 刘林, 林山驰, 李相国, 冯敏, 许亮. 基于改进的U-Net网络模型的叶片病害检测. 液晶与显示, 2024, 39: 1138–1144.

Liu L, Lin S C, Li X G, Feng M, Xu L. Leaf disease detection based on improved U-Net network model. Chin J Liquid Crystal Displays, 2024, 39: 1138–1144 (in Chinese with English abstract).

[16] 洪俐, 刘涛, 孙成明, 左示敏, 严长杰. 基于改进U-Net++的水稻病害图像分割研究. 江苏农业科学, 2024, 52(12): 201–208.

Hong L, Liu T, Sun C M, Zuo S M, Yan C J. Study on rice disease image segmentation based on improved U-NET ++ model. Jiangsu Agric Sci,

2024, 52(12): 201–208 (in Chinese with English abstract).

[17] Zhou Z W, Rahman Siddiquee M M, Tajbakhsh N, Liang J M. UNet++: a nested U-Net architecture for medical image segmentation. In: Zhou Z W, Rahman Siddiquee M M, Tajbakhsh N, Liang J M, eds. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. Cham: Springer International Publishing, 2018. pp 3–11.

[18] Wang G W, Wang J W, Wang J X, Sun Y D. Grape leaf disease classification combined with U-net++ network and threshold segmentation. Comput Intell Neurosci, 2022, 2022: 1042737.

[19] Xiong X, Yu L J, Yang W N, Liu M, Jiang N, Wu D, Chen G X, Xiong L Z, Liu K D, Liu Q. A high-throughput stereo-imaging system for quantifying rape leaf traits during the seedling stage. Plant Methods, 2017, 13: 7.

[20] Liang X Y, Xu X C, Wang Z W, He L, Zhang K Q, Liang B, Ye J L, Shi J W, Wu X, Dai M Q, et al. StomataScorer: a portable and high-throughput leaf stomata trait scorer combined with deep learning and an improved CV model. Plant Biotechnol J, 2022, 20: 577–591.

[21] Wu W, Liu T, Zhou P, Yang T L, Li C Y, Zhong X C, Sun C M, Liu S P, Guo W S. Image analysis-based recognition and quantification of grain number per panicle in rice. Plant Methods, 2019, 15: 122.

[22] 梁丽秀. 基于MATLAB的水稻表层根系图像分割算法研究. 华中农业大学硕士学位论文, 湖北武汉, 2018.

Liang L X. Research on Image Segmentation Algorithm of Rice Surface Root System Based on MATLAB. MS Thesis of Huazhong Agricultural

University, Wuhan, Hubei, China, 2018 (in Chinese with English abstract).

[23] Wu X, Feng H, Wu D, Yan S J, Zhang P, Wang W B, Zhang J, Ye J L, Dai G X, Fan Y, et al. Using high-throughput multiple optical phenotyping to decipher the genetic architecture of maize drought tolerance. Genome Biol, 2021, 22: 185.

[24] Hu J, Shen L, Albanie S, Sun G, Wu E H. Squeeze-and-excitation networks. In: Hu J, Shen L, Albanie S, Sun G, Wu E H, eds. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Computer Vision Foundation, 2018. pp 7132–7141.

[25] Misra D, Nalamada T, Arasanipalai A U, Hou Q B. Rotate to attend: convolutional triplet attention module. In: Misra D, Nalamada T, Arasanipalai A U, Hou Q B, eds. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Computer Vision Foundation, 2021. pp 3139–3148.

[26] Yu T, Li X, Cai Y F, Sun M M, Li P. S²-MLPv2: improved spatial-shift MLP architecture for vision. Tex: Cornell University, 2021 [2021-08-02]. https://arxiv.org/abs/2108.01072v1.

[27] 温静. 通过广度优先搜索有向图重组规则次序. 湖北大学学报(自然科学版), 2004, 26(4): 294–296.

Wen J. Reordering rule sequences by breadth-first searching directed graph. J Hubei Univ (Nat Sci Edn), 2004, 26(4): 294–296 (in Chinese with

English abstract).

[28] 刘云飞, 张俊然. 深度神经网络学习率策略研究进展. 控制与决策, 2023, 38: 2444–2460.

Liu Y F, Zhang J R. Research advances in deep neural networks learning rate strategies. Control Decision, 2023, 38: 2444–2460 (in Chinese with

English absract).

[29] 蒋文斌, 彭晶, 叶阁焰. 深度学习自适应学习率算法研究. 华中科技大学学报(自然科学版), 2019, 47(5): 79–83.

Jiang W B, Peng J, Ye G Y. Research on adaptive learning rate algorithm in deep learning. J Huazhong Univ Sci Technol (Nat Sci Edn), 2019, 47(5): 79–83 (in Chinese with English abstract).

[30] Berman M, Triki A R, Blaschko M B. The lovasz-softmax loss: a tractable surrogate for the optimization of the intersection-over-union measure in neural networks. In: Berman M, Triki A R, Blaschko M B, eds. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Computer Vision Foundation, 2018. pp 4413–4421.

[31] 谢俊, 杨超宇. 基于改进YOLOv5s的高压环境穿戴绝缘用具检测算法. 重庆科技学院学报(自然科学版), 2023, 25(6): 59–64.

Xie J, Yang C Y. Study on algorithm for detecting insulated tools in high voltage environments based on YOLOv5s. J Chongqing Univ Sci Technol (Nat Sci Edn), 2023, 25(6): 59–64 (in Chinese with English abstract).

[32] Li C X, Liu X Y, Li W Y, Wang C, Liu H Y, Liu Y F, Chen Z, Yuan Y X. U-KAN makes strong backbone for medical image segmentation and generation. Tex: Cornell University, 2024 [2024-08-22]. https://arxiv.org/abs/2406.02918v3.

[33] Lan L B, Cai P Z, Jiang L, Liu X J, Li Y M, Zhang Y D. BRAU-net++: U-shaped hybrid CNN-transformer network for medical image segmentation. Tex: Cornell University, 2024 [2024-09-30]. https://arxiv.org/abs/2401.00722v2.

[34] Cao H, Wang Y Y, Chen J, Jiang D S, Zhang X P, Tian Q, Wang M N. Swin-unet: unet-like pure transformer for medical image segmentation. In: Cao H, Wang Y Y, Chen J, Jiang D S, Zhang X P, Tian Q, Wang M N, eds. European conference on computer vision. Cham: Springer Nature Switzerland, 2023. pp 205–218.

[35] Ibtehaz N, Kihara D. ACC-UNet: a completely convolutional UNet model for the 2020s. In: Ibtehaz N, Kihara D, eds. International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer Nature Switzerland, 2023. pp 692–702.

[36] Minervini M, Fischbach A, Scharr H, Tsaftaris S A. Finely-grained annotated datasets for image-based plant phenotyping. Pattern Recognit Lett, 2016, 81: 80–89.

[37] Zhang X H, Huang C L, Wu D, Qiao F, Li W Q, Duan L F, Wang K, Xiao Y J, Chen G X, Liu Q, et al. High-throughput phenotyping and QTL mapping reveals the genetic architecture of maize plant growth. Plant Physiol, 2017, 173: 1554–1564.

[38] 赵亚凤, 刘晓璐, 王冬冬, 王孟雪, 宋文华, 胡峻峰. 基于改进U-Net的根系表型参数测量系统. 森林工程, 2024, 40(4): 127–136.

Zhao Y F, Liu X L, Wang D D, Wang M X, Song W H, Hu J F. Root phenotypic parameter measurement system based on improved U-Net. For Engin, 2024, 40(4): 127–136 (in Chinese with English abstract).

[39] 陈才武. 甘蓝型油菜pol TCMS温度感受器基因突变体的创制. 华中农业大学硕士学位论文, 湖北武汉, 2023.

Chen C W. Creation of Thermoreceptor Genes Mutants for pol TCMS in Brassica napus L. MS Thesis of Huazhong Agricultural University, Wuhan, Hubei, China, 2023 (in Chinese with English abstract)

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed