Drone hyper-spectral imaging was used in this study to evaluate the severity of wheat stem rust disease in a disease trial containing 960 plots. Quadratic data analysis and random forest classifier, decision tree classification, and support vector machine were applied to select the wavelengths and spectral vegetation indices (SVIs). The trial plots were divided into four levels based on ground truth: class 0 (healthy, scored 0), class 1 (mildly diseased, scored 1-15), class 2 (moderately diseased, scored 16-34) and class 3 (severely diseased, scored 35+). The RFC method achieved the highest classification accuracy, which was 85% for overall classification. The RFC method with selected SVIs, and the accuracy ranged between 82%-96%. Green NDVI, Photochemical Reflectance Index, Red-Edge Vegetation Stress Index and Chlorophyll Green were selected from 14 SVIs. It is possible to build a new inexpensive multispectral imaging system for stem rust disease detection.

The monitoring evapotranspiration over crops field can provide the information of crop water requirement for Agricultural Optimization, since soil water evaporation reflects the water content in the soils, which determined the water available to crops through crops root systems; meanwhile the transpiration reflects the water conditions in plants. Our recent developed hyperspectral imaging technique at near short wave infrared had demonstrated the ability to monitoring the vapor fluctuations with sensitivity at under 70-micrometer perceptible atmospheric water. This makes the hyperspectral imaging technique is suitable for monitoring the evapotranspiration over vegetation fields. With optimal optical-mechanics design, the current ruggedized setup is portable with the total mass under 20 kg, while keeping the adaptive spatial sampling rate around 8000~ 35,000, and adaptive spectral sampling rate over 20~40 in the wavelength range of 1100nm to 1300nm. A series of observations from different vegetation fields were carried out under different weather conditions. Beside the Evapotranspiration information, the classification of the fields through machine learning based on 3d convolutional neural networks are achieved. The evapotranspiration information from the snap shot imaging spectrometer are consistent with the theoretical modelling results based on MODTRAN.

Extreme weather events could lead to variations in agricultural materials produced within the same field. This can complicate decision-making for growers. The deviations in nutritional contents of feed materials were investigated using standard techniques. Statistically significant differences (P < 0.05) were observed in the feed materials within and among factories. A line-scan hyperspectral imaging system was used to develop rapid and non-destructive global measurement models for smart decision-making with an accuracy of up to 86%. The models show potential for precision feeding leading to high productivity and the reduction in emissions from overfeeding and inefficiency.

As one of the primary agricultural commodities in Korea, Chinese cabbage is susceptible to disease infections. The plants which exposed to a high moisture are easily infected by downy mildew disease. The disease is identified by irregular yellow-tan spots appearing on the upper leaf surface, leading to cell damage thus degrading the product quality. An early detection system to identify and treat the disease would be essential to prevent disease occurrence and reduce the plant damage caused by the disease. Hyperspectral imaging, as one of the non-destructive evaluation methods, has recently become more popular due to its capability to capture a wide range of light spectrum. It is sensitive enough to detect slight chemical difference within the plant. UAV-based hyperspectral system offers high-throughput plant phenotyping with abundant resources of data. A preliminary experiment has shown spectral differences between diseased and healthy cabbage leaves. Based on hyperspectral image data, the detection system employs a convolutional neural network (CNN) that extracts spectral and spatial features to detect the disease and its location. A 3D CNN architecture will be used in this study to further exploit the spectral variance and accurately detect the disease.

This study was designed to develop accurate spatial thermal and vegetative indices maps of corn field and derive precise crop water stress index using thermal imagery. UAV (Matrice 100) mounted, FLIR Vue Pro thermal imaging camera and MicaSense Rededge multispectral camera were used to collect aerial imagery over a corn field divided into different irrigation levels. Weather data such as ambient air temperature, relative humidity, solar radiation, wind speed were also recorded to establish crop water stress baselines. The collected images were stitched and converted into orthomosaics using Agisoft. Canopy temperature, crop water stress index (CWSI), NDVI and NDRE maps were generated from orthomosaics using Arc map. The real time ground data in terms of soil moisture, crop canopy temperature and leaf water potential was also collected. Results showed that imagery based crop water stress was strongly co-related with different vegetative indices and real-time ground data.

Breeders and researchers widely use remote sensing by UAS as a high-throughput phenotyping tool. However, the experiment condition, such as time of day, season, crop type, plant structure, and illumination setting, can influence the measured data, impact the analytical outcomes and lead to uncertain conclusions. In this study, we investigated the effects of Sun-camera geometry on plants' reflectance measured by multispectral cameras. For this purpose, over two years, we collected an extensive dataset of UAS-based imagery from several citrus, almond, and grape orchards in various environmental conditions. We measured and documented the significant effect of sun-camera geometry on the reflectance data. We realized that an off-nadir view angle of 25 degrees within the field of view in a single image could lead to a 50% variation in measured reflectance compared to the nadir view. The results indicate that the object's reflectivity could be a function of its position within an image frame independent of its inherent spectral characteristics. The variations emerge from direct solar radiations (as opposed to diffused radiations) and climax in clear sky conditions around the solar noon. We proposed a model that estimates and correct the reflectance variations due to sun-camera geometry. The results showed that the proposed model could estimate reflectance from other view angles by r2 of up to 0.88 depending on the spectral band, crop type, and a few other variables. Comparing these results with the data simulated by the 4Sail model showed a reasonable consistency in the outcome. This study's findings help determine the optimum flight time as a function of the camera's FOV, date, and the site's latitude, which will facilitate the correction of the direct solar radiation effect and assist in avoiding unreliable data collection.

We developed a mast-mounted hyperspectral imaging polarimeter (HIP) that images a corn field across multiple diurnal cycles throughout a growing season. Using the polarization data, we present preliminary results demonstrating the potential to use polarization to de-couple light reflected from the surface versus light scattered from the tissues, thus enabling time of day, solar incidence angle, and viewing angle to be reduced as confounding factors for the spectral measurement. Polarization correction is achieved through training neural networks and by creating a scattering model of corn leaves by measuring the Bidirectional Reflectance Distribution Function (BRDF).