By knowing how much fruit is on the trees before harvest, growers can better plan for labor needs and harvesting.
Harvesting accounts for about 30 percent of overall citrus production costs, according to a news release.
Wonsuk "Daniel" Lee, a University of Florida agricultural and biological engineering professor, used the algorithm to find immature citrus in photos taken under different light conditions. The algorithm also was used to identify fruit hidden by leaves or branches.
He and his colleagues found about 80 percent of immature fruit using the system.
The trial was conducted at a research grove on the UF campus in Gainesville.
Using a camera, two of Lee's former students took 240 photos of fruit.
Growers typically estimate crop yields based on years of experience. The U.S. Department of Agriculture also issues monthly crop estimates during the season based on surveying representative groves and counting and measuring fruit on selected branches.
“This gives growers a more accurate rate than just guessing,” Lee said in the release. But he did note that his method isn’t quite ready to be used to estimate yield for an entire grove.
When that day comes, he says growers will benefit: “If you know the exact yield, you can predict the price.”
In smaller groves, it's feasible to photograph every tree. But in larger groves, operators would photograph representative trees.
As the next step, Lee and a graduate student are working to develop a self-propelled machine vision system for growers.