AI Case Study

University of Montana is testing an app with which beekeepers will be able to try to find out what is troubling a colony by listening to the buzz

The university of Montana was in the final stage of testing an app, as of March 2018, developed to detect if bees are suffering from a number of maladies based on the sound they are making. Based on a database of sound recordings of healthy and unhealthy bees, an artificial neural network is trained to match bee sounds to those associated with certain hive problems.

Industry

Basic Materials

Agriculture

Project Overview

"The resulting app, which is called Bee Health Guru, is being produced by Bee Alert Technology, a company spun out from the university.

For such an idea to work, it is necessary to attribute specific bee ailments to particular sounds. To do that, the university tapped into its worldwide network of beekeepers to find colonies that were known to suffer from only one problem, and to obtain sound recordings of bees in those colonies. The sounds that bees make come from their beating wings (although movements by other parts of their bodies may also be involved). Having built up a database of sounds, an artificial neural network, a form of machine learning used for pattern recognition, was employed to help build algorithms that can match bee sounds to those associated with certain hive problems.

To check on the health of a colony of bees it is usually necessary to open the hive, a procedure which involves using smoke to pacify the bees. That is a time- consuming process for commercial beekeeping operations, some of which may have several thousand colonies to take care of. With the app, all a beekeeper need do is to hold their smartphone near to the hive’s entrance for 30 seconds while it analyses the sound of the bees. The app then lists any health problems which it detects.

Seven different disorders will at first be checked, says David Firth, a team member who is helping to bring the app to market. These include the presence of hive beetle, a serious honeybee pest, parasitic mites and “foulbrood”, a bacterial infection which can destroy bee colonies.

With the permission of users, data from the app can be shared with the researchers, who plan thereby to update the software to detect other diseases and problems, says Dr Firth. This could include exposure to pesticides, in particular a group called neonicotinoids which are suspected of harming honeybees (pesticide producers reject such claims). Finally, if all works to plan, bees will get to have their say about the things that cause them harm."

Reported Results

Proof of concept; results not yet available

Technology

"Having built up a database of sounds, an artificial neural network, a form of machine learning used for pattern recognition, was employed to help build algorithms that can match bee sounds to those associated with certain hive problems."

Function

R And D

Core Research And Development

Background

Those afflictions might provide an indication of an impending Colony Collapse Disorder (CCD), a mysterious syndrome that has plagued beekeepers in North America and Europe. Unlike a natural swarm, in which a large group of worker bees leave with their queen to form a new colony, CCD involves bees suddenly disappearing for no obvious reason, leaving their queen behind. Although recent reports suggest there has been a reduction in bee die-offs, according to some estimates 10m hives in America alone were wiped out by CCD from 2006 to 2013. Besides hitting honey production, this can also hinder the pollination of certain crops.

Benefits

Data

sound recordings of bees