Final Report – 2021-2023 Funding Cycles (two year project)
Principal Investigator: Manoj Karkee, director of the Center for Precision and Automated Agricultural Systems, Washington State University, 24106 N. Bunn Rd, Prosser, Washington
509-786-9208
manoj.karkee@wsu.edu
Summary: Automated crop estimation would be a valuable tool for the Washington wine industry. Efforts to develop automated crop estimation tools have been stymied by the vine’s canopy that prevents cameras from seeing individual clusters. This research furthered earlier work by WSU to develop a mobile-app for crop estimation for the Washington wine industry based on a simple and low-cost approach using smartphone camera and GPS systems and cloud computing platform for image processing. In that work, there was low correlation between the number of berries and cluster weight of visible and total clusters, which suggested a weak correlation between the image count and the actual count due to the variable shape of the clusters and different degree of occlusion of berry clusters within the model. However, in this expanded research and further work on the model, the difference between the total estimated yield and the actual yield was 10 percent, which is promising.
An added focus of the research was to develop a model to detect lag phase in vineyards. The growth trends of individual berries of sample clusters of Merlot and Chardonnay exhibited similar patterns of steady growth, followed by stunted growth and then another period of rapid growth. Data was collected over the growing season on ten clusters for each variety, representing two years in Merlot and one year in Chardonnay. Manual measurement of berry size was also collected for sample berries over the same period. A mathematical equation was used to model the berry growth pattern, which achieved relatively high correlation of 0.96 for Merlot and Chardonnay when 2022 data was used.
WSU’s grape crop estimation app can detect, locate and count clusters within a vineyard. Work is underway to calibrate estimated berry count and size against manual measurements and develop a correlation model to estimate cluster weights using the app. The app is under beta testing. Lag phase detection and estimation of total yield at the field level will be integrated into the app in the future.
Download the report above to learn more.