Search

Artificial Intelligence (AI) has advanced rapidly in the past two decades. Internet of Things (IoT) technology has advanced rapidly during the last decade. Merging these two technologies has immense potential in several industries, including agriculture. We have identified several research gaps in utilizing IoT technology in agriculture. One problem was the digital divide between rural, unconnected, or limited connected areas and urban areas for utilizing images for decision-making, which has advanced with the growth of AI. Another area for improvement was the farmers' demotivation to use in-situ soil moisture sensors for irrigation decision-making due to inherited installation difficulties. As Nebraska has a unique agricultural industry, including 8.6 million irrigated croplands accounting for 14.8% of all irrigated croplands in the United States and a significant amount of under-connected farmlands, we decided to address these two pressing issues. Chapters three and five discuss the design and development of the AICropCAM platform. AICropCAM can extract useful information such as crop type, canopy coverage, and pest pressure and transmit it to cloud servers through low-power, long-range communication mediums using deep learning models. This platform can reduce the digital divide in crop monitoring digital solutions. Chapter four discusses an effort to predict the soil water properties, including soil water content, soil water deficit, and remaining water content, through machine learning models that depend on the above canopy sensors and real-time weather data. The current models were trained using 2022 and 2023 data, showing promising prediction accuracy above 95%. However, these models must be tested in the long run for commercial use. At the top level, the case studies discussed in this dissertation will help reduce the rural-urban gap and digital divide in agriculture. Additionally, the proposed AICropCAM and the above-ground sensor use for soil water property estimation can increase pivot irrigation systems' crop monitoring and improve irrigation decision-making tools. Ultimately, the case studies in this dissertation conclude that the integration of AI and IoT can advance the crop monitoring practices used for precision agriculture in mid-western irrigated cropping systems. Advisor: Yufeng Ge

Multispectral imaging systems on satellite, aerial, and ground platforms are used commonly to monitor in-field crops in precision agriculture by farmers and researchers. Limited spatial and temporal resolution and weather dependence of the data collection are two main disadvantages of these methods. In-field sensor networks can continuously monitor environmental and plant physiological parameters by leveraging low-power computation and long-range communication technologies. We built and tested a novel sensor network equipped with soil moisture, multispectral and RGB imaging sensors in an experimental soybean field at Eastern Nebraska Research and Extension Center, NE, USA. 10 down-looking and 1 up-looking sensor node were set up at the experimental site in the 2020 growing season. Each down-looking sensor node utilized an 18-channel multispectral sensor, a soil moisture sensor, and an RGB imager to collect canopy reflectance, soil volumetric water content (VWC), and canopy images every 20 minutes. The up-looking sensor node measured the solar radiation using a multi-spectral sensor every 10 minutes. The setup allowed us to calculate the spectral reflectance of the soybean canopy under changing weather conditions. The sensor nodes were solar-powered and integrated into a Low Power, wide area network (LoRaWAN) through a LoRa gateway, which was connected to the internet via Wi-Fi. Captured images were saved on SD cards while other parameters were uploaded to cloud data storage for real-time processing and visualization. The result shows that the sensor network can plot canopy Normalized difference vegetation index (NDVI) and soil VWC continuously throughout the growing season. NDVI values of the irrigated and rainfed soybean plots showed a significant difference. The coefficient of determination or the R squared value was 0.8701 between the GreenSeeker Handheld Crop Sensor and the IoT enabled sensor node NDVI value. This research verified that NDVI value is not constant throughout the day. Daily NDVI variation has two peaks between 10.00 -11.00 am and 1:00 - 2:00 pm. This sensor network could help users to estimate the crop growth parameters and irrigation requirements in a real-time fashion. Further, the diurnal NDVI tracking with an in-field NDVI sensor node has the potential to improve the NDVI value-based variable-rate fertigation unit efficiency improvement. Advisor Yufeng Ge