What is the PotatoPi?
The PotatoPi is a prototype Raspberry Pi computer vision system that I am developing. It uses a Raspberry Pi 3 and a Pi camera in order to automatically count and sort potatoes, or other items. The system is prototyped in python, but will be later converted to C. It’s intended to be a product identification and tracking system that is as cheap and modular as possible.
Why make it?
I am designing the PotatoPi in order to increase productivity and introduce product tracking abilities within a potato warehouse.
How does it work?
The PotatoPi uses Python3 and OpenCV in order to capture and process images from a Pi Camera module. The images that are captured by the Pi Camera module are processed locally on the Raspberry Pi 3. The package imutils, made by Adrian Rosebrock, is used in order to easily multi-thread the PiCamera module for more efficient processing.
Images are captured and processed at a rate of approximately 30 frames per second. The images are transformed after being captured using OpenCV functions, and objects are detected based on previously input parameters. The detected objects are added to an accumulator.
It’s possible to access the PotatoPi system remotely using a VNC program in order to monitor and control everything.
I intend to add multiple levels of classification to the project. Potatoes will be identified by color and variety, and defects will be identified and isolated. Mutiple product types will be incorporated into the system, to allow for realtime tracking of multiple classes of products.
Other potential additions to the system are determining dimensions and area of detected objects, as well as velocity or acceleration.
Currently the project is in early prototyping, and requires the user to set up and start the program through the graphical desktop. I am going to make it so that everything will launch automatically when the system is turned on.
Questions? Contact me directly:
[contact-form-7 id=”40″ title=”Contact form 1″]