Hyperspectral Image Analysis for Nondestructive Estimation of Crop Nutrient Composition

The ability to efficiently manage nutrients is one of the most important components of sustainable agriculture since it affects crop yields, resource utilization, and the long-term health of the environment. Nutrient diagnostics has proven to be successful in providing accurate results in laboratories. However, due to their nature, laboratory-based diagnostics are typically time-consuming, destructive, and less than ideal for both large-scale and real-time monitoring. Hyperspectral imaging (HSI) offers an alternative to laboratory-based nutrient diagnostics in terms of being nondestructive, while having the potential to accurately determine the chemical composition of a crop's nutrient content. HSI uses a combination of two technologies to create a detailed spectral response across hundreds of narrow, adjacent bands of spectral wavelength (from the visible through the short-wave infrared [SWIR] wavelengths, 400–2500 nm). These spectral signatures contain valuable biochemical and physiological information related to the nutrient status of a plant. Different nutrient deficiencies produce different spectral responses; for example, nitrogen deficiency impacts the red-edge portion of the spectrum (680–750 nm), phosphorus impacts the blue-green portion of the spectrum, and potassium impacts the absorption of SWIR radiation, depending on the moisture levels within the leaf. Micronutrient deficiencies (such as iron, zinc, and manganese) also produce detectable spectral changes. The versatility of HSI technology enables the assessment of nutrients to occur at various spatial scales, from leaf-level precision to larger regional scales using a variety of platforms (including ground-based sensors, UAVs, aircraft, and satellites). Data preprocessing techniques must be used to improve the quality of the data before it can be used for analysis purposes. Advanced computational models can be used to establish a relationship between the spectral reflectance and the nutrient concentration in plant tissues with a high degree of predictive accuracy. Combining HSI with other complementary technologies (i.e., LiDAR, thermal, and RGB) allows for the enhancement of the diagnostic capability through multimodal data integration. While there continue to be many obstacles (cost of sensor, complexity of data, etc.) to overcome to utilize this technology for real-time and cost-effective, and scalable applications, innovations such as spectral libraries, transfer learning, and edge computing are allowing for the development of HSI-based tools for real-time, cost-effective, and scalable applications. Ultimately, HSI has the potential to be a powerful tool for precision agriculture and promote sustainable and climate-smart nutrient management.