BREAD PHENO: High-Throughput Phenotyping with Smart Phones. #phenoApps

未来几十年，粮食和营养安全将是一个巨大的挑战。全球人口预计将增加到90亿以上，粮食需求将增长50%以上。目前，全世界有20亿人生活在贫困中，主要依靠发展中国家的自给自足农业。虽然贫穷和粮食不安全是一个复杂的问题，但培育适应气候变化、高产和营养丰富的植物品种是改善粮食安全、增加收入和经济福利的关键部分。为了应对这一挑战，需要创新的方法来加快改良植物品种的开发。在植物育种和遗传学中，需要对植物特性进行精确的测量，以准确地确定重要基因的作用，并识别和选择最有前途的候选植物品种。这一领域的技术开发有限，特别是在田间试验中测量的性状，大多数测量仍然是手工进行和记录的。该项目将开发用于测量植物特性的移动应用程序(App)，这些应用程序可以部署在廉价且随时可用的移动设备上。通过与木薯和小麦育种者合作进行初步测试和部署，将能够迅速传播和广泛使用。初中生和高中生也将参与测试和使用这些应用程序来探索植物生长和测量植物特征。为世界各地数以千计的植物育种者配备快速测量和分析重要植物特性的工具，将为加快开发改良植物品种奠定基础，最终将提高发展中国家小农及其家庭的生产率、粮食安全、营养和收入。在过去的十年里，基因组数据的可用性呈爆炸式增长，而收集表型的方法却取得了很小的进步。这导致了连接基因型和表型的数据集的显著不平衡，并突出了表型鉴定仍然是植物育种计划中的主要瓶颈。该项目将通过将同步地面真实表型测量和成像与移动技术相结合，推动3D图形和建模、数据挖掘和深度学习领域的发展。在现场手册(www.meatgentics.org/field-book)取得成功的基础上，将开发和部署用户友好的现场高通量表型分析(HTP)移动应用程序。该项目将融合图像处理和机器视觉方面的新进展，通过现有的繁育网络提供移动应用程序。将开发新的图像分析算法来模拟和提取植物表型。1)通过世界各地的育种合作者进行实时田间测试，以及2)中学生和高中生使用这些应用程序来探索基因控制下的植物生长和数量差异，这将有助于一个强大的发展管道。为了确保在多种作物上立即、广泛地部署和发挥功能，将参与木薯和小麦育种网络，提供一套不同的目标植物表型、环境、育种计划和工作菌种。通过将研究方案的数据与地面真相繁殖者的知识相结合，该项目将为收集训练集奠定基础，这些训练集随后可用于利用深度学习提取和量化复杂的表型。智能手机和平板电脑的开源应用程序将包括软件和文档，这样用户就能够了解如何使用这些应用程序。应用程序将通过在线应用程序商店(Windows Store、iTunes App Store、Google Play)、项目网站和合作植物育种网络进行分发。生成的源代码将托管在拥有GNU通用公共许可证(GPL)开放源码许可证的公共GitHub库中。

Food and nutritional security will be a grand challenge in the coming decades. The global population is expected to increase to over 9 billion and food demand will grow by more than 50%. Currently, there are 2 billion people worldwide living in poverty, mostly relying on subsistence agriculture in developing countries. While poverty and food insecurity is a complex issue, the development of improved climate-resilient, high yielding and nutritious plant varieties is a critical part of improving food security, increasing income and economic welfare. To address this challenge, innovative approaches are needed to speed up the development of improved plant varieties. In plant breeding and genetics, precise measurements of plant characteristics are needed to accurately determine the effect of important genes and to identify and select the most promising candidate plant varieties. There has been limited technology development in this area, particularly for traits measured in field trials where most measurements are still taken and recorded by hand. This project will develop mobile applications (apps) for measuring plant traits that can be deployed on inexpensive and readily available mobile devices. Initial testing and deployment through collaboration with cassava and wheat breeders will enable rapid dissemination and broad usability. Middle-school and high-school students will also be engaged to test and use the apps to explore plant growth and measure plant traits. Equipping thousands of plant breeders around the world with tools for rapid measurement and analysis of important plant traits will provide the foundation for accelerated development of improved plant varieties that will ultimately result in increased productivity, food security, nutrition and income of smallholder farmers and their families in developing countries. Over the past decade, the availability of genomic data has exploded while the methods to collect phenotypes have made minimal advancements. This has led to a dramatic imbalance in data sets connecting genotype to phenotype and highlights phenotyping as the remaining major bottleneck in plant breeding programs. This project will advance the field of 3D graphics and modeling, data mining and deep learning through integration of simultaneous ground truth phenotypic measurements and imaging with mobile technology. Building on the success of Field Book (www.wheatgenetics.org/field-book), user-friendly mobile apps for field-based high-throughput phenotyping (HTP) will be developed and deployed. This project will converge novel advances in image processing and machine vision to deliver mobile apps through established breeder networks. Novel image analysis algorithms will be developed to model and extract plant phenotypes. A robust development pipeline will be assisted by 1) real-time field testing through breeding collaborators around the world and 2) middle-school and high-school students using the apps to explore plant growth and quantitative differences under genetic control. To ensure both immediate, broad deployment and functionality on a diverse set of crops, breeder networks for cassava and wheat will be engaged, providing a diverse set of target plant phenotypes, environments, breeding programs and working cultures. By combining data from research programs with ground truth breeder knowledge, this project will lay the foundation for collecting training sets that can subsequently be used to extract and quantify complex phenotypes using deep learning. Open-source apps for smartphones and tablets will consist of both software and documentation so that users will be able to understand how to use the apps. Apps will be distributed through online app stores (Windows Store, iTunes App Store, Google Play), through project websites, and via collaborative plant breeding networks. The resulting source code will be hosted in a public GitHub repository with a GNU General Public License (GPL) open-source license.