Wheat stripe rust is a significant disease affecting wheat growth, often referred to as the “cancer of wheat”. Early and accurate detection of stripe rust is crucial for enabling crop managers to implement effective control measures. Hyperspectral remote sensing methods for crop disease detection have gained significant attention. However, commonly used spectral bands or spectral indices (SIs) from hyperspectral data often fail to capture the subtle changes associated with the early stages of crop diseases accurately. In this study, we propose a method for early detection of wheat stripe rust by combining pigments and SIs retrieved from UAV hyperspectral imagery. We acquired hyperspectral images of wheat stripe rust at 7, 16, and 23 days post-inoculation (DPI) using a UHD 185 hyperspectral sensor (450–950 nm) mounted on an S1000 hexacopter UAV. Pigments, including chlorophylls (Cab), carotenoids (Car), anthocyanins, Cab/Car, and 11 pigment-related SIs, were extracted from UAV hyperspectral images using radiative transfer modeling. The early detection model for wheat stripe rust was developed using these parameters and machine learning algorithms. The results indicated selected pigments and SIs effectively distinguished stripe rust-infected wheat from healthy wheat at 7, 16, and 23 DPI. Models that combine pigments and SIs (PSIMs) perform better than those relying solely on SIs (SIMs) or pigments (PMs). Notably, the RF-based PSIM achieved overall accuracies of 78.1 % and 81.3 % during the asymptomatic (7 DPI) and minimally symptomatic (16 DPI) phases of disease, respectively. Additionally, the pigments in the PSIM contributed more significantly than the SIs, highlighting the importance of pigments in the early detection of stripe rust. Overall, the method combining pigments and spectral indices proposed in this study effectively enhances the early detection of wheat stripe rust and offers valuable insights into the early detection of other crop diseases.

中文翻译：

---

通过整合从无人机高光谱图像中量化的颜料和颜料相关光谱指数，改进了小麦条锈病的早期检测


小麦条锈病是影响小麦生长的重要疾病，通常被称为“小麦癌”。及早准确检测条锈病对于使作物管理者能够实施有效的控制措施至关重要。用于作物病害检测的高光谱遥感方法受到了广泛关注。然而，来自高光谱数据的常用光谱波段或光谱指数 （SI） 往往无法准确捕捉与作物病害早期阶段相关的细微变化。在这项研究中，我们提出了一种通过结合从无人机高光谱图像中检索的色素和 SI 来早期检测小麦条锈病的方法。我们使用安装在 S7 六旋翼无人机上的 UHD 185 高光谱传感器 （450-950 nm） 在接种后 16 天、 23 天 （DPI） 获取了小麦条锈病的高光谱图像 （DPI）。使用辐射传输建模从无人机高光谱图像中提取色素，包括叶绿素 （Cab）、类胡萝卜素 （Car）、花青素、Cab/Car 和 11 个与色素相关的 SI。小麦条锈病的早期检测模型就是使用这些参数和机器学习算法开发的。结果表明，在 7、16 和 23 DPI 下，选定的色素和 SIs 有效地区分了条锈病感染的小麦和健康小麦。将颜料和 SI （PSIM） 相结合的模型比仅依赖 SI （SIM） 或颜料 （PM） 的模型表现更好。值得注意的是，基于 RF 的 PSIM 在疾病的无症状 （7 DPI） 和轻微症状 （16 DPI） 阶段分别实现了 78.1% 和 81.3% 的总体准确率。此外，PSIM 中的色素比 SI 贡献更大，突出了颜料在条锈病早期检测中的重要性。 总体而言，本研究中提出的结合色素和光谱指数的方法有效地增强了小麦条锈病的早期检测，并为其他作物病害的早期检测提供了有价值的见解。