Trash detection system for a citrus canopy shake and catch harvester using machine vision

摘要

Automatic estimation of the amount of trash, such as branches and leaves, collected by a mechanical citrus harvester during harvesting eliminates problems in the processing plants with handling diseased leaves and fruit. A machine vision system was developed to estimate the amount of trash collected by a citrus canopy shake and catch harvester by acquiring and analyzing the images of the harvested materials passing through the harvester's conveyor belt Over 27,000 images were acquired from a commercial citrus grove during harvesting in Ft Basinger, Florida on May 9-10, 2008. These images were processed and the trash objects were identified from a representative set of 180 images. The weight of the trash object was estimated using a calibration set of branch and leaf samples obtained from the mechanical harvester. The trash weight estimates were correlated with GPS data to form a geo-referenced map of the trash gathered. The results from the trash detection system can be used to come up with betterways of filtering out the trash from the harvester so that most of the trash could be disposed of in the field during harvesting.