SoylentNews is people
SoylentNews
from the She-thinks-my-Raspberry-Pi's-sexy dept.
In farming regions that receive limited moisture, fallow phases are used to maximize crop yield potential with special attention given to weed removal. When fields are laying fallow, it is very important to remove the weeds that are trying to establish themselves. When the density of the weeds is low, it is not efficient to spray the whole field with herbicide, so proprietary sensor-based spray booms are used to identify and spray herbicide directly onto the weeds. Researchers from the University of Sydney have developed an open-source, low-cost and image-based device for weed detection in an effort to give this technology wider availability. The demonstration system uses a Raspberry Pi 4 with a Raspberry Pi HQ Camera and 6-mm focal length lens. The Python code and detailed instructions reside in the OWL github repository.
OWL represents a novel opportunity for community-driven development of weed recognition capability using existing ‘off-the-shelf’ hardware and simple yet effective image-based algorithms. The combination of the OWL device, supporting documentation and repository create a channel for practical education of key image-based weed detection and actuation concepts for growers and the wider weed control community. The topic is of particular importance now, given the emergence of image-based in-crop weed recognition technologies. OWL has been designed as a community-focused educational platform that will grow over time with initial baseline validation performed in the present research.
Reference:
Guy Coleman, William Salter, and Michael Walsh. OpenWeedLocator (OWL): an open-source, low-cost device for fallow weed detection. [open] Sci Rep 12, 170 (2022). (DOI: 10.1038/s41598-021-03858-9)
(Score: 3, Interesting) by richtopia on Thursday January 13 2022, @09:08PM (6 children)
The Github page is more readable than the paper, located here: https://github.com/geezacoleman/OpenWeedLocator [github.com]
It is quite well documented. It looks like you drag a trailer over your field with downward cameras to inspect the ground from about 1m away. They perform some image recognition to identify the weeds poking up.
I'm a gardener who will pretend to follow farming practices on my little plot. I really appreciate this technique; moving towards no-till techniques and removing herbicides makes weed control more labour intensive. This OWL uses focused herbicide specifically on the identified weed. You'll lose your organic certification but in my opinion this is a reasonable compromise. Maybe mechanical removal could be automated in the future.
(Score: 3, Interesting) by Thexalon on Thursday January 13 2022, @09:49PM (1 child)
I'll admit that my favorite discovery in gardening my small plot was how awesome a hoe is for weed control. Just the difference between standing up with that rather than on my knees weeding was huge. There was still some hand-weeding involved when the bad guys are right next to the plants I want to keep, but once you get good with it you can get a lot without disturbing too much of the good stuff.
The only thing that stops a bad guy with a compiler is a good guy with a compiler.
(Score: 0) by Anonymous Coward on Saturday January 15 2022, @05:18PM
When I use hoes, I don't usually mind the "getting down" bit at all.
(Score: 3, Informative) by hubie on Friday January 14 2022, @02:31AM (1 child)
FarmBot [youtube.com] doesn't use herbicides, but it doesn't scale very well. :)
(Score: 0) by Anonymous Coward on Friday January 14 2022, @12:54PM
Per the video, Farmbot scans for weeds just getting started, and pushes them into the soil to kill them. No chemicals or lasers needed. It's the automated version of tending a small plot daily--catch weeds before they become big enough to be a problem.
(Score: 0) by Anonymous Coward on Friday January 14 2022, @07:13AM (1 child)
There is a company testing and/or manufacturing a laser based weed killing machine.
(Score: 0) by Anonymous Coward on Friday January 14 2022, @08:07AM
Pictures or it didn't happen