In this work, basil plants were fertilized with 0, 2.5 mM and 10 mM nitrogen (with different NO3−/NH4+ ratios), and then monitored using a low-power technique based on an optical leaf meter and a low-cost RGB sensor interfaced with an Arduino UNO board. The study aimed to investigate possible relationships between the concentration of some plant compounds (i.e., leaf chlorophyll and flavonoids content) and the nitrogen balance index, with the output data of a low-cost RGB sensor to indicate its capability in discriminating among different levels of nutrition. The data obtained underwent univariate and multivariate analysis. The univariate data analysis showed that the low-cost RGB sensor readings followed the development of the plants according to the varying applications of nitrogen. The multivariate analysis of the data showed that the indices related to plant metabolic efficiency and leaf colour were those most affected by the nitrogen levels of the solutions used. The comparison of the discrimination powers of the systems showed that both systems achieved comparable discrimination performances (85.0% and 89.4%) for plants supplied with 0 mM nitrogen solution. However, at increasing levels of nitrogen, the RGB sensor performed worse than the optical leaf meter (− 15.8% and − 8.6% for the 2.5 and 10 mM N treatments). The effect of the NO3−/NH4+ ratio could hardly be distinguished (except for the total chlorophyll resulting from the optical leaf meter readings). More data is, however, necessary to create a more robust model for future implementation of the application of such a sensor.

中文翻译：

---

应用低成本 RGB 传感器在中试水平检测罗勒（Ocimum basilicum L.）营养状况

在这项工作中，罗勒植物用 0、2.5 mM 和 10 mM 氮（具有不同的 NO3-/NH4+ 比率）施肥，然后使用基于光学叶计和低成本 RGB 传感器接口的低功率技术进行监测使用 Arduino UNO 板。该研究旨在研究某些植物化合物的浓度（即叶绿素和黄酮类化合物的含量）与氮平衡指数之间的可能关系，并利用低成本 RGB 传感器的输出数据来表明其区分不同水平的能力。营养。获得的数据进行了单变量和多变量分析。单变量数据分析表明，低成本 RGB 传感器读数根据氮的不同应用跟踪植物的发展。数据的多变量分析表明，与植物代谢效率和叶色相关的指标受所用溶液中氮含量的影响最大。系统区分能力的比较表明，对于供应 0 mM 氮溶液的植物，两种系统都实现了可比的区分性能（85.0% 和 89.4%）。然而，随着氮含量的增加，RGB 传感器的性能比光学叶计差（对于 2.5 和 10 mM N 处理分别为 - 15.8% 和 - 8.6%）。几乎无法区分 NO3-/NH4+ 比率的影响（除了由光学叶计读数得出的总叶绿素）。然而，需要更多的数据来创建一个更强大的模型，以便将来实现这种传感器的应用。