UAV Subsystem for Computer Vision-Assisted Soybean Herbicide Injury Quantification
Herbicide injury on soybean crops greatly reduces crop yield. Current soybean injury evaluation methods are visual based and there are UAV based methods being developed. Visual based injury evaluation results could be unreliable because every field personnel might have different judgment on the level of injury. In addition, visual evaluation may require traveling the entire field with a vehicle which requires extra time, and the evaluation data may lack spatial resolution. UAV based methods require tedious data collection in large quantities and processing of this data requires a computer with fast processing speeds. The processing of remote sensing data may take hours sometime more than a day to develop a map that could provide another visual estimate of the herbicide injury level. The objective of this research is to develop a lightweight low-cost on-board computer system to evaluate herbicide injury on soybean crops using computer vision methods. UAV assisted computer vision-based instantaneous pixel quantification and data geotagging of injured tissues of the soybeans are being investigated. The novel UAV based herbicide soybean injury subsystem will provide instantaneous accurate data collection. The same data then could be used by the same computer to control another actuator or system to treat injured plants if a treatment is possible.
Preliminary Analysis of Blackberry Blossom Quantification with UAV Computer Vision
Blackberry phenotyping studies include quantification of blackberry blooms for various purposes. Visual counting of the blooms is widely used among researchers and plant breeders, and it is the most accepted method. This method could be time consuming when considering large size experiment sites and requirement of repetition of bloom counting. The blackberry bloom density estimation is made based on the field personal’s visual judgment and usually scaled from one to ten, one indicating low density and 10 indicating high density. UAV remote sensing and GIS based estimations have been tested. However, the blooms are small and cannot be seen in the orthomosaic generated from UAV images at standard UAV flight altitudes above 30 m. That requires near ground altitude flights increasing the number of pictures to be processed to have orthomosaic from the entire area. Increased number of pictures to be processed at very low altitude increases the time that is required to create orthomosaic. The objective of this research is to develop a computer vision-based blossom density measurement UAV subsystem. The UAV system could fly autonomously at very low altitudes between 1 m to 12 m above blackberry canopy, measure blackberry density to generate blossom index, and record the geotagged data in its memory to be mapped later in GIS software. This method would provide high resolution accurate blackberry blossom density measurements with little effort from the field personnel.
Computer Vision-Assisted Weed Pressure Quantification for Variable Rate UAV Spraying
Weed management is one of the most important agricultural practices that is crucial for crop production. Worldwide, weed competition causes severe yield reduction specifically in wheat, soybean, rice, maize, cotton, and potato. Postemergence and preharvest weed management require appropriate timing because a delay in weeding may increase weed population for the following year. Continuous use of non-residual herbicide with same sites of action encourages herbicide resistance and motivates the development of new resistant weed populations. Reducing the dependency on herbicide usage while maintaining a timely weed management practice is needed to reduce costs, minimize environmental impact, and reduce herbicide carryover. The objective of this research is to develop a subsystem for an Unmanned Aerial Vehicle (UAV) that can instantaneously adjust spray rate while measuring weed density in soybean fields. To achieve a variable rate spraying from a UAV, preliminary studies have been made to determine efficient spraying altitude and on-the-go weed pressure estimations in soybean fields.
UAV-Assisted Water Quality Monitoring in Lakes
Water quality monitoring activities in surface waters require continues repeated water sampling events to evaluate water quality. Considering the large number of waterbodies, and their varying sizes, it’s difficult and costly to implement sensor nodes in each water body. In addition, traveling to each site on a single water body and other water body requires extra time. Most of the time, some waterbodies are left unmeasured, for example their distance to operation base, health concerns due to algal bloom or E. Coli. Algal blooms may produce toxins resulting in livestock death creating economic burden on producers. The objective of this research is to investigate UAV applications in water quality monitoring and develop novel methods for UAV-assisted water sampling, in-situ measurements, and remote sensing.