Insect wingbeats will help quantify biodiversity
Date:
February 22, 2022
Source:
University of Copenhagen - Faculty of Science
Summary:
Insect populations are plummeting worldwide, with major consequences
for our ecosystems and without us quite knowing why. A new AI
method is set to help monitor and catalog insect biodiversity,
which until now has been quite challenging.
FULL STORY ========================================================================== Insect populations are plummeting worldwide, with major consequences
for our ecosystems and without us quite knowing why. A new AI method
from the University of Copenhagen is set to help monitor and catalogue
insect biodiversity, which until now has been quite challenging.
========================================================================== Insects are vital as plant pollinators, as a food source for a wide
variety of animals and as decomposers of dead material in nature. But in
recent decades, they have been struggling. It is estimated that 40 percent
of insect species are in decline and a third of them are endangered.
Therefore, it is more important than ever to monitor insect biodiversity,
so as to understand their decline and hopefully help them out. So far,
this task has been difficult and resource-intensive. In part, this is
due to the fact that insects are small and very dynamic. Furthermore, scientific researchers and public agencies need to set up traps, capture insects and study them under the microscope.
To overcome these hurdles, University of Copenhagen researchers have
developed a method that uses the data obtained from an infrared sensor to recognize and detect the wingbeats of individual insects. The AI method
is based on unsupervised machine learning -- where the algorithms can
group insects belonging to the same species without any human input. The results from this method could provide information about the diversity
of insect species in a natural space without anyone needing to catch
and count the critters by hand.
"Our method makes it much easier to keep track of how insect populations
are evolving. There has been a huge loss of insect biomass in recent
years. But until we know exactly why insects are in decline, it is
difficult to develop the right solutions. This is where our method can contribute new and important knowledge," states PhD student Klas Rydhmer
of the Department of Geosciences and Natural Resource Management at
UCPH's Faculty of Science, who helped develop the method.
Advanced artificial intelligence The researchers have already developed
an algorithm that identifies pests in agricultural fields. But instead of identifying insects as pests, the researchers have been able to develop
this new algorithm to identify and count various insect populations in
nature based on the measurements obtained from the sensor.
==========================================================================
"The sensor is a bit like the wildlife surveillance cameras used to
monitor the movements of larger animals in nature. But instead of snapping
a photo, the sensor measures insects that have has flown into the light
source. The algorithm then uses the insect's wingbeat to identify them
into different groups," explains Assistant Professor Raghavendra Selvan
of the Department of Computer Science, who led the development of the artificial intelligence used in the sensor.
The algorithm distinguishes insects by their silhouettes when their
wings are folded out, as it is only then that their physical differences
become most apparent. It then compares the silhouettes of different
insect recordings, and puts similar silhouettes into the same group
which can then be used to determine the insect that most likely flew
through the light beam.
Prototype to be released in spring When insects emerge in full force
come spring, scientists will be using the initial prototype to venture
out into nature and collect real-world data.
Until now, researchers have tested the algorithm and artificial
intelligence using a large image database of insects recordings obtained
in controlled conditions and some real-world data, where results have
been promising.
==========================================================================
"We will test the sensor in different landscapes, including heathland,
forests and agricultural areas, to see how it works out in the real
world. But also, to feed the algorithm more data, so that it can become
even more accurate," says Raghavendra Selvan.
According to the researchers, their invention makes it possible to monitor
many geographical areas more thoroughly than has been possible in the
past. At the same time, the invention makes it less resource-intensive to
keep a close eye on insects, which make up 80 percent of all terrestrial
animal species.
"Today, it is impossible to afford the kind of monitoring needed to
gain a more precise overview of how our insects are doing. This sensor
only needs humans to place it out in the wild. Once there, it begins
collecting data on local insect populations," concludes Klas Rydhmer.
Background:
* Insects are the largest, most diverse group of described animal
species
on Earth. They make up about 80% of all terrestrial animal species
on the planet.
* It is the first time that this artificial intelligence method,
known as
Variational Auto Encoder (VAE), is being used to take inventory
of insect biodiversity.
* Using an optical signal from an infrared sensor, the algorithm is
able to
decode insects flying through a light beam.
========================================================================== Story Source: Materials provided by University_of_Copenhagen_-_Faculty_of_Science. Note: Content may be
edited for style and length.
========================================================================== Journal Reference:
1. Klas Rydhmer, Raghavendra Selvan. Dynamic b-VAEs for quantifying
biodiversity by clustering optically recorded insect
signals. Ecological Informatics, 2021; 66: 101456 DOI:
10.1016/j.ecoinf.2021.101456 ==========================================================================
Link to news story:
https://www.sciencedaily.com/releases/2022/02/220222135250.htm
--- up 11 weeks, 3 days, 7 hours, 14 minutes
* Origin: -=> Castle Rock BBS <=- Now Husky HPT Powered! (1:317/3)