A robot-enabled, data-driven machine vision tool for nitrogen diagnosis of arable soils

在减少投入和减少对环境影响的情况下提高作物生产力是全球粮食生产面临的一项重大挑战。在所有农民控制的投入因子中，氮(N)对作物生长的影响仅次于水和可耕地是许多种植系统中氮的主要来源。优化施氮可以提高作物产量，提高土壤肥力。另一方面，高氮投入对农民来说代价高昂，并导致植物生物多样性减少、自然生态系统污染和强效温室气体一氧化二氮排放增加。英国农民目前在化肥上的支出为13.45亿英镑。目前，农民使用了过高的施氮量，因为他们不知道哪些土地上氮过量、最佳或缺乏。土壤氮的准确检测对种植系统的经济和环境可持续性至关重要。目前测定土壤氮的做法成本高、劳动密集、耗时长，因此高氮投入很常见。在终端用户需求的驱动下，学术界、工业界和农民之间的这一创新跨学科项目将首次共同开发一种具有成本效益、非破坏性、机器人驱动、数据驱动的机器视觉解决方案，利用颠覆性技术（机器人技术、人工智能/计算机视觉/大数据分析）、农业科学来自动检测耕地土壤的氮水平，能够1)使用具有3D成像传感的移动机器人自动收集数据；2)基于人工智能/计算机视觉/大数据分析的作物氮和土壤氮的自动智能诊断，利用作物和/或覆盖作物作为土壤氮值的定量生物指标来获取氮。这利用了已知的植物对氮富集的反应，模型由本研究中得到的精确关系参数化。传感器系统和机器视觉/数据分析与自主机器人平台的集成为食品生产中新的测量和基于机器视觉的任务提供了重要的机会，否则这些任务是无法获得和无法实现的。这种精准农业解决方案将通过减少氮素投入、提高农场盈利能力来改变粮食生产，并通过提供作物氮素和土壤氮素的早期检测、提供氮素利用效率和土壤质量评估的准确信息，通过减少一氧化二氮的排放，为实现净零排放做出贡献。它将使英国精准农业技术处于新兴产业的前沿，并推动英国的经济增长。

Increasing crop productivity with reduced inputs and lower impacts on the environment is a major challenge for global food production. Among all farmer-controlled input factors, Nitrogen (N) has the second-largest impact on crop growth after water and arable soils are a predominant source of N in many cropping systems. Optimal N fertilisation can increase crop production and enhance soil fertility. On the other hand, high N inputs are costly for farmers and result in reductions in plant biodiversity, pollution of natural ecosystems and increases in emissions of the potent greenhouse gas, nitrous oxide. Current spend on fertilisers by UK farmers is £1.345bn. At present excessively high N fertiliser rates are used by farmers because they are not aware of the areas of land where N is excessive, optimal or deficient. Accurate detection of soil N is crucial for the economic and environmental sustainability of cropping systems. The current practices in determining soil N is costly, labour intensive and time consuming and so high N inputs are common.For the first time, driven by the end user needs, this innovative interdisciplinary project between academia, industry and farmers will co-develop a cost-effective, non-destructive, robot enabled, data driven, machine vision solution by harnessing disruptive technologies (Robotics, AI/Computer Vision/Big Data Analytics), agricultural science to automatically detect nitrogen levels of arable soils, capable of 1) automated data collection using a mobile robot with 3D imaging sensing; 2) automated intelligent diagnosis of both crop N and soil N based on AI/Computer Vision/big data analytics which derives N using the crop and/or cover crop(s) as a quantitative bioindicator of soil N values. This utilises known plant responses to N enrichment with models parameterised by precise relationships derived in this study. The integration of sensor systems and machine vision/data analytics with autonomous robotic platforms offer significant opportunities for new measurements and machine vision-based tasks in food production that would otherwise be unobtainable and unachievable. This precision agriculture solution will transform food production by reducing N inputs, increasing farm profitability and contribute to net-zero emission by reducing emissions of nitrous oxide through offering early detection of both crop N and soil N, providing accurate information on nitrogen N use efficiency and soil quality assessment. It will position UK precision agriculture technologies at the forefront of new industries and drive economic growth in UK.