Color Segmentation for Vegetables Quality Control

The goal of this work consists in a software capable in processing images of agricultural products to verify their quality using information about color and shape. Input images are taken by a RGB camera in 256x256 pixel format and then converted in HSI system, in order to permit a better performing segmentation algorithm.

Picture Here

Our segmentation algorithm recognizes big objects like tomatoes, peppers or artichokes, medium objects like stems, and sometime also small objects like little dents and insect stings on the surface of vegetables.

Picture Here "Not Good"

We developed an interactive software to help the user in building easily the right setup for each kind of vegetable, where specifying color, kind of shape, wanted particulars and unwanted ones.

Loaded a setup file, the software should become able to find a vegetable in a greenhouse ambient and to produce its report based on the similarity of the input vegetable with the relating setup. Shape algorithms are based on simple geometrical measuring like area, perimeter, and inertia axis.

Picture Here "Good"

Reports and Articles

Back to past projects@LIRA