Many studies have shown that hyperspectral measurements can help monitor crop health status, such as water stress, nutrition stress, pest stress, etc. However, applications of hyperspectral cameras or scanners are still very limited in precision agriculture. The resolution of satellite hyperspectral images is too low to provide the information in the desired scale. The resolution of either field spectrometer or aerial hyperspectral cameras is fairly high, but their cost is too high to be afforded by growers. In this study, we are interested in if the flow-cost hyperspectral scanner SCIO can serve as a crop monitoring tool to provide crop health information for decision support. In an onion test site, there were three irrigation levels and four types of soil amendment, randomly assigned to 36 plots with three replicates for each treatment combination. Each month, three onion plant samples were collected from the test site and fresh weight, dry weight, root length, shoot length etc. were measured for each plant. Meanwhile, three spectral measurements were made for each leaf of the sample plant using both a field spectrometer and a hyperspectral scanner. We applied dimension reduction methods to extract low-dimension features. Based on the data set of these features and their labels, several classifiers were built to infer the field treatment of onions. Tests on validation dataset (25 percent of the total measurements) showed that this low-cost hyperspectral scanner is a promising tool for crop water stress monitoring, though its performance is worse than the field spectrometer Apogee. The traditional field spectrometer yields the best accuracy as high as above 80%, whereas the best accuracy of SCIO is around 50%.
In the last decade, technologies of unmanned aerial vehicles (UAVs) and small imaging sensors have achieved a significant improvement in terms of equipment cost, operation cost and image quality. These low-cost platforms provide flexible access to high resolution visible and multispectral images. As a result, many studies have been conducted regarding the applications in precision agriculture, such as water stress detection, nutrient status detection, yield prediction, etc. Different from traditional satellite low-resolution images, high-resolution UAVbased images allow much more freedom in image post-processing. For example, the very first procedure in post-processing is pixel classification, or image segmentation for extracting region of interest(ROI). With the very high resolution, it becomes possible to classify pixels from a UAV-based image, yet it is still a challenge to conduct pixel classification using traditional remote sensing features such as vegetation indices (VIs), especially considering various changes during the growing season such as light intensity, crop size, crop color etc. Thanks to the development of deep learning technologies, it provides a general framework to solve this problem. In this study, we proposed to use deep learning methods to conduct image segmentation. We created our data set of pomegranate trees by flying an off-shelf commercial camera at 30 meters above the ground around noon, during the whole growing season from the beginning of April to the middle of October 2017. We then trained and tested two convolutional network based methods U-Net and Mask R-CNN using this data set. Finally, we compared their performances with our dataset aerial images of pomegranate trees. [Tiebiao- add a sentence to summarize the findings and their implications to precision agriculture]
Irrigated potato production in sandy soils can be impacted by low nitrogen (N) and water retention in the soil. A field study was conducted to use canopy spectral reflectance as a primary means to characterize N fertilizer rates and soil texture variations as growth and yield limiting factors in potato. A hand-held 16-band spectral radiometer was used to obtain reflectance readings of the potato canopies. Reflectance measurements were made in field plots that received four rates of N or in four areas where the soil textures were different. At later stages of plant growth, canopy reflectance in the 760 to 1000 nm spectral range was consistently higher in plots that received higher rates of N or in areas where the soil contained higher clay and silt fractions. Russet Burbank potatoes, with increasing rate of N fertilizer, showed a decreasing trend in total tuber yield and an increasing trend in percent of tubers with weight exceeding 170 g. Canopy reflectance was inversely related to tuber yield or size for Russet Burbank potatoes when soil texture was the only variable.