Date fruit segmentation for automated harvesting

In robotic harvesting for date fruit, the robot first recognizes the type and maturity of date bunches and determines the harvesting decision: ready to harvest or not. Then, the segmentation takes place to localize each date bunch and locate the position to cut the stalk. This is needed to move the arm of the robot to harvest the bunch. Fruit segmentation is also important to estimate the weight of date bunches and for yield prediction.

Two solutions are proposed, the first solution utilizes superpixel clustering and local binary patterns (LBP), and the second solution employs deep learning-based semantic segmentation.  

The proposed solutions are prepared for publication in an article.