A Multi-Modal System For Yield Prediction in Citrus Trees

One of the thrusts of the Integrated Automation for Sustainable Specialty Crops Farming project funded by USDA is focused on the development of advanced sensing and modeling technologies for predicting citrus yield. This paper describes the approaches developed and evaluated so far, along with their estimated prediction accuracy. The described capabilities include systems for modeling tree canopies using mobile ground based sensors (laser range finders, stereo and color cameras mounted on vehicles). These sensors are used to generate canopy features including tree volume, tree canopy density and orange counts. The different sources of information are combined through a regression model. Preliminary experimental results from a commercial citrus grove in Florida are presented.