Use of Machine Vision to Sort Tomato Seedlings for Grafting Robot

Abstract

Using a machine vision with backlighting LEDs, a tomato-seedling grading and sorting algorithm was developed for a fully automatic grafting robot. A UXGA camera, a blue color backlighting device and light filtering devices were selected to acquire seedling images. The developed algorithm was used to determine the bending, nodes of leaves and stem diameter of seedlings from their images and then to grade and sort them as the initial task of the grafting robot. Results showed that the sorting success rate was 97% and the rest 3% was failed as the target portions of the images of seedlings were covered by irregular arrangement or bended leaves from all sides.