This is a mini-project for the image processing course of the FHGR B.Sc. Mobile Robotics, implemented by Dominic Eicher and Ralph Ursprung.
This project analyses pictures of bananas to try and identify some information about them (e.g. their ripeness). Note that we will currently only support Cavendish Bananas (the standard banana) as others may exhibit other properties (e.g. cooking bananas are green even when ripe).
Note that this implementation does not rely on any machine learning algorithms and instead employs traditional image processing steps.
The project consists of a library implementing the actual functionality and two applications, one feeding the library with live pictures from an attached camera and one feeding it static images (mainly for manual testing). Additionally, the 2D polyfitting library has been split into its own library as it is separate from the rest.
To build this project you will need:
- A modern C++ compiler supporting C++23
- CMake incl. CTest - this might well come included with your favourite IDE
- vcpkg
Here is an example of the application processing an image with two bananas on it:
Additionally, the following output is generated on the console, offering further information:
found 2 banana(s) in the picture
Banana #0:
y = 1968.704538 -2.818917 * x +0.001434 * x^2
Rotation = -6.42 degrees
Mean curvature = 0.21 1/cm (corresponds to a circle with radius = 4.76 cm)
Length along center line = 9.59 cm
ripeness = 103 %
Banana #1:
y = 493.926139 -0.944471 * x +0.001329 * x^2
Rotation = 81.35 degrees
Mean curvature = 0.20 1/cm (corresponds to a circle with radius = 4.92 cm)
Length along center line = 9.46 cm
ripeness = 104 %
As this is purely an educational project there's no need for others to include it in their commercial works.
Accordingly, this is licensed under the GNU General Public License v3.0 or later (SPDX: GPL-3.0-or-later).
See LICENSE for the full license text.