|Zap A Mosquito With A Pi|
|Written by Harry Fairhead|
|Sunday, 06 June 2021|
and a laser of course. It isn't a new idea but at last someone has given it a go and made some results available. You might be surprised what the biggest problem is.
I really don't need to make the case for zapping mozzies. Even if you ignore the 700,000 deaths attributed to their bite every year, they are simply annoying. There have been some high-level promises to develop a laser-based system, notably Nathan Myhrvold who, since he retired from his job as chief technical officer of Microsoft, had promised a range of amazing technology, but to date none have claimed success and been converted into anything remotely commercial.
Rakhmatulin Ildar from the South Ural State University decided to do something about it. He put together a system based on a Raspberry Pi that can detect a mosquito using computer vision. It then uses a simple galvanometer-based pointing system to aim the laser and fire.
The setup made use of Python and OpenCV and, to make things simpler and controllable, the mosquitoes were kept in a box 30cm from the device. So we really are shooting mosquitoes in a barrel - -well box... A pair of cameras was used to provide depth location.
It would seem that a neural network is the obvious choice, but after looking at several pretrained networks the idea was rejected because of the limited processing power and the second or so it took to compute a result. Clearly there is not much point in knowing where a mosquito was one second earlier.
In the end four different methods, all based on traditional non-neural network image processing were tried: Color, tracking, cascade operators, frame difference and optical flow. All but optical flow achieved a reasonable level of tracking accuracy and the simplest, color detection, worked well.
The depressing part of the result is that, even after hitting a mosquito, only about 50% of the insects were killed! Clearly we need something more powerful than a1W laser used or some focusing technique.
1 - box with mosquitoes, 2 - pi cameras, 3 - Raspberry PI3, 4 - laser, 5 - galvanometer, 6 - laser beam
Clearly more work is needed, but it is also clear that you don't need millions in funding to make progress. Open MozzyZap anyone?
or email your comment to: email@example.com
|Last Updated ( Sunday, 06 June 2021 )|