In order to develop fully-automatic grafting robot for tomato seedlings, the machine vision system for sorting grafting seedling and inspecting grafted seedling was studied. In this paper, the devices and different lighting (Front lighting, Backlight, and both Front lighting and Backlight) of machine vision system were optimized and selected. The result shows that Blue Backlight without filter device is best for both sorting and inspecting seedlings. Then the images of sorting and inspecting seedlings were acquired using the selected machine vision system and processed by the program. Experiment results show the success rate on processing the images of sorting seedlings can reach 96% and even the gap of 0.2 mm between scion and rootstock can be found easily when inspecting grafted seedling. Copyright © 2013 American Scientific Publishers.