无人机遥感技术在植物表型分析中的新突破：IHUP软件平台的开发与应用
研究背景
随着传感器小型化和成本降低，无人机（UAV）低空遥感技术在田间植物表型实验中的应用日益广泛。然而，跨学科知识的需要和复杂的工作流程严重阻碍了研究人员从多源和多时相的无人机图像中提取地块级表型数据。这一挑战限制了植物表型研究的效率和准确性，亟需一种集成化、自动化的解决方案。
研究内容
为了解决上述问题，研究团队开发了集成高通量通用表型分析（IHUP）软件平台。该平台包含四个主要功能模块：预处理、数据提取、数据管理和数据分析。通过集成和自动化处理，简化了需要复杂跨学科知识的数据提取和分析过程。用户可以在图形用户界面中计算图像特征信息、结构特征和植被指数（VIs），这些是形态和生化特征的指标。此外，为了满足不同作物的数据需求，提取方法如VI计算公式可以进行定制。 为了验证软件的组成和性能，研究团队进行了与水稻干旱表型监测相关的实验。结合水稻叶片卷曲分数预测模型，从多阶段连续监测数据中高效提取了叶片卷曲分数、植株高度、VIs、鲜重和干旱重量等特征。尽管图像处理在地块裁剪过程中对处理效率有显著影响，但软件在大多数应用案例中能够以每分钟约500个地块的速度提取特征。
研究意义
IHUP软件平台的创新之处在于其用户友好的图形用户界面和可定制或集成各种特征提取算法的接口，这显著降低了非专家用户的门槛。该软件有望显著加速无人机表型实验中的数据生产，为植物表型研究提供了一个高效、准确的工具。此外，IHUP的开发和应用不仅提高了数据处理的效率，也为植物科学研究提供了一个新的视角，有助于推动农业科技的进步。 通过这项研究，我们可以看到技术进步和创新如何推动农业科学的发展，特别是在提高作物产量和适应性方面，IHUP软件平台将发挥重要作用。
Abrstract
With the threshold for crop growth data collection having been markedly decreased by sensor miniaturization and cost reduction,unmanned aerial vehicle(UAV)-based low-altitude remote sensing has shown remarkable advantages in field phenotyping experiments.However,the requirement of interdisciplinary knowledge and the complexity of the workflow have seriously hindered researchers from extracting plot-level phenotypic data from multisource and multitemporal UAV images.To address these challenges,we developed the Integrated High-Throughput Universal Phenotyping(IHUP)software as a data producer and study accelerator that included 4 functional modules:preprocessing,data extraction,data management,and data analysis.Data extraction and analysis requiring complex and multidisciplinary knowledge were simplified through integrated and automated processing.Within a graphical user interface,users can compute image feature information,structural traits,and vegetation indices(Vis),which are indicators of morphological and biochemical traits,in an integrated and high-throughput manner.To fulfill data requirements for different crops,extraction methods such as VI calculation formulae can be customized.To demonstrate and test the composition and performance of the software,we conducted case-related rice drought phenotype monitoring experiments.In combination with a rice leaf rolling score predictive model,leaf rolling score,plant height,VIs,fresh weight,and drought weight were efficiently extracted from multiphase continuous monitoring data.Despite the significant impact of image processing during plot clipping on processing efficiency,the software can extract traits from approximately 500 plots/min in most application cases.The software offers a user-friendly graphical user interface and interfaces for customizing or integrating various feature extraction algorithms,thereby significantly reducing barriers for nonexperts.It holds the promise of significantly accelerating data production in UAV phenotyping experiments.
文章来源：https://spj.science.org/doi/10.34133/plantphenomics.0164
相关来源：https://www.lunlunapp.com/newsDetails/c1dd18bed2539ca42b283ce09027aadf
