Progress 06/01/15 to 01/31/17
Target Audience: Farmers/growers Crop consultants, agronomists Application contractors/Weed control businesses Agricultural co-ops and "channel companies" (e.g., WinField, CPS) Farm equipment manufacturers Herbicide/GMO conglomerates This technology has potential impact at all levels of US agriculture. Local farmers/growers will have available a more precise method of weed control, allowing them to simultaneously reduce herbicide use and better control HR weed populations, thus getting higher yields while saving money on inputs. This technology will "trickle up" within the agricultural industry. Crop consultants/agronomists will have more options in their prescriptions and recommendations. Application contractors/weed control businesses and regional agricultural companies and co-ops can offer additional services.Herbicideand GMO conglomerates can use previously un-marketed chemicals which were either too expensive to use in broadcast spraying or damaging to the crop plants. Changes/Problems: Deviation from research schedule: To provide sufficient time to acquire seeds from herbicide-resistant weed populations and cultivate them for the leaf-specific herbicide application study, USDA granted a 1-year, no-cost extension due for the project, bringing the total project time from 8 months to 20 months. Small deviation from research goal: Although we originally proposed to use only the visible light spectrum for imaging and analysis, we also began incorporating simultaneous imaging in the near infra-red (NIR) spectrum. This opened the possibility of enhancing our data set with variations of the normalized difference vegetation index (NDVI) calculation. Unexpected outcome: Our image processing speeds have advanced orders of magnitude beyond those originally anticipated due to advances in both our hardware design and software algorithms. As a result, a slow-moving autonomous platform is no longer required to accomplish leaf-specific herbicide application. Instead, existing, human-driven farm vehicles can be used, moving at common tractor speeds. Unexpected outcome: We learned that for some types of weeds (e.g., marestail), applying a herbicide droplet to the central grow point can be more effective than applying a droplet to multiple leaves. This finding may influence our target selection moving forward. Unexpected outcome: We now intend to consider on-label usage of currently approved herbicides in addition to original plan of off-label use of unapproved chemicals. This outcome is not science- or engineering-related, but a result of our commercialization research: that the combined political, legal, financial, strategic partnering, IP, and time scale barriers to off-label chemical use are larger than anticipated. What opportunities for training and professional development has the project provided?
How have the results been disseminated to communities of interest?
What do you plan to do during the next reporting period to accomplish the goals?
What was accomplished under these goals?
Project Description Concurrent Solutions, LLC (CS-LLC) is developing the science and technology for controlling weeds, including herbicide-resistant (HR) weeds, by placing herbicides exclusively onto an individual weed leaf (or leaves) in the presence of crop plants in a field, without spraying and without impacting nearby crop plants. The project addresses two issues. First, the proliferation of HR weeds in recent decades, which is at least partially attributable to accelerated natural selection pressure due to broadcast spraying of non-selective herbicides on GMO crops. The proliferation of HR weeds has had the effect of reducing crop yields while increasing the costs of weed control. Second, providing an alternative or complement to broadcast spraying would also reduce herbicide drift beyond fields, runoff into water bodies, and the amount of herbicide residuals in the food supply, while giving farmers more latitude in seed selection to optimize yields. These project goals support 4 of the USDA's strategic goals. Activities CS-LLC is uniquely qualified to develop the proposed technology for the agricultural domain, based on their years of experience developing sensing and kinetic strike technologies for the DoD. During the Phase I effort our team established the feasibility of our proposed technology by focusing on a specific crop (soybeans) and specific weeds of concern to farmers (Palmer amaranth and marestail). CS-LLC personnel collected field image data of both weeds and crops and used these images to develop effective and efficient software algorithms capable of identifying and differentiating crop leaves and weed leaves in real-time from a moving platform. Simultaneously, our weed expert at the University of Kentucky performed a novel dose-response study on HR strains of weeds of concern to demonstrate that our leaf-specific application method is capable of controlling weeds. Finally, CS-LLC personnel engineered and constructed prototype mobile imaging equipment and precision herbicide ejectors capable of addressing the kinetic challenges of our proposed goals. The results of this Phase I work has demonstrated the promise of our technological approach to leaf-specific weed control. Impact This technology has potential impact at all levels of US agriculture as follows. Researchers, regulators, crop consultants, and agronomists will learn the benefits of the approach and begin recommending it to growers. Farmers will begin choosing to use the technology to better control HR weeds and increase their yields, thus creating a market demand. In response, application contractors, weed control businesses, farm co-ops, and regional agricultural companies will begin to offer new services using large-scale leaf-specific herbicide application. These changes will drive farm equipment manufacturers to innovate and manufacture new applicator products and allow herbicide manufacturers and GMO companies to use previously un-marketed chemicals which were either too expensive to use in broadcast spraying or damaging to the crop plants. The end result will be new conditions for US agriculture, such as a diminishing rather than increasing HR weed problem, more options for seed selection and herbicide treatments, and smaller amounts of herbicides in the environment and food supply. Accomplishments We identified 4 main objectives to achieve our project goals, relating to machine vision, targeting/delivery, application rates, and dosage/formulation. Both our goals and objectives are described above. The methods, data, and results are summarized below, by task: Objective 1 - Machine Vision: We collected field images at all growth stages in both the visible and near infrared spectrums. We used the images to develop machine vision software algorithms capable of differentiating between soybean plants, Palmer amaranth, and marestail. We based these algorithms on related software we built capable of identifying black grass stalks in wheat fields. These algorithms are feasible because our cameras resolve 3 pixels per millimeter, which is orders of magnitude more detail than competing imaging systems. This resolution is even theoretically enough to identify differences between standard Palmer and HR strains Palmer strains because it can pick up leaf traits such as petiole length, smoothness, and chevron colorations. Objective 2 - Targeting & Delivery: We constructed a number of experimental apparatus, including a 40 foot long greenhouse, a 24 foot long gantry with computer-controlled sled, and two prototype high-precision herbicide droplet ejectors. We conducted experiments in platform motion, targeting, splatter, and wind effects. Accuracy and splatter tests using our new herbicide ejector prototypes were significant improvements over the versions we built in 2004. Our new prototype applicator can dispense 1-10 microliter herbicide droplets within 1-10 milliseconds (at up to 261 shots per second) and hit a ½ inch target from a moving platform traveling 2 feet above the target. The splatter from such a shot will leave 99.9% of the herbicide on the leaf, causing minimal impact to adjacent leaves and no impact to leaves more than 1 inch away. Objective 3 - Application Rates: We conducted experiments using our imaging system, gantry, and herbicide ejector prototypes. We plugged the results into analytical models that factor in variables including, but not limited to, target density, processing speed, targeting speed, firing rate, platform motion, herbicide droplet time-of-flight, and wind deflection. We designed and built specialized computer hardware optimized to run our machine vision and targeting algorithms in the least amount of time possible. Our model shows that, with a sufficiently large array of herbicide ejector nozzles, it would be practical using current technology to tow a leaf-specific herbicide ejector system at speeds of up to 8 mph. Objective 4 - Dosage & Formulation: Our weed expert at University of Kentucky selected 5 systemic herbicides for the dose-response study after considering sites/modes of action from 12 WSSA herbicide groups. The novel study focused on off-label herbicide usage at high concentrations applied to weeds up to 6 inches tall. Rather than spraying plants, concentrated herbicides were directly applied to selected weed leaves (for Palmer amaranth) and grow points (for marestail) at precise doses ranging from 1 to 10 microliters using a micro pipette. Glyphosate emerged as the most effective herbicide, capable of killing all but the largest plants in the study using just 5 microliters applied to 1-2 leaves or the growing point. Conclusion In successfully completing the four project objectives, we have demonstrated the feasibility of our leaf-specific herbicide application approach to precision weed control in row crops. Our approach is ideal for weeds up to 8 inches in height, post crop emergence and prior to canopy closure. We can use computer vision to automatically identify weeds, target them, and apply discrete droplets of herbicide to them without impacting the adjacent crops. The tiny amounts of active ingredient required (micrograms) applied to ~1-3 leaves of each plant represent a 90-99% reduction in herbicide use for full-field weed control. The farm equipment involved can be driven through a field at speeds of up to 8 mph. We believe that any one of our results is significant enough to warrant further work. Taken in aggregate, our results may represent a new paradigm for precision weed control.