Using Artificial Intelligence to Protect Bees - A Solution from this Year's "Digital Innovation Lab"
29 July 2022
Photo: pixabay/foto-suju
Written by Anna Priebe (published on July 27 in the UHH Newsroom)
How can mite infestation of bee colonies be detected at an early stage? An interdisciplinary team of students has developed a technology for automated acoustic monitoring. To do so, they used the infrastructure of a new co-working space at the university. Nabil Basharat and Kevin Kraus report on their joint creative work.
Species protection and insect mortality are very topical issues. How did you come up with the idea of tackling the problems with artificial intelligence methods?
Kevin Kraus: The starting point for the search for topics in the project format "Digital Innovation Lab" (led by Prof. Dr. Jan Recker) was the United Nations' 17 Sustainable Development Goals. While brainstorming, I told my girlfriend about a start-up that uses image recognition to evaluate the bees flying in and out of a beehive and the amount of pollen they bring back with them, and she said that her mother had had bees for a while. But they died of the varroa mite - a type of mite that attacks hives and is a major reason for bee deaths worldwide. Again, there are approaches that use image recognition to check for infestations. We wanted to further automate this technique, but it has proven very difficult to implement.
How does your alternative approach work?
Kraus: We conducted further research and came across the acoustics of the bees as an indicator of how healthy the colony is. It turned out, in fact, that you can tell from the buzzing, for example, whether a queen is present, but also whether there is an infection with the Varroa mite. As a result, we decided to use this monitoring approach.
Our prototype consists of a small microcontroller, for example a chip that contains various control elements and is powered by solar power. It's effectively a mini-PC that records sounds and examines them with a machine learning model. We trained the model with recordings of bee sounds and tested it again and again. This means that the investigation takes place directly in the device on site and only the result of the classification has to be transmitted at the end.
What have been the challenges?
Kraus: Machine learning depends heavily on the quality of the data you use to train the artificial intelligence. Unfortunately, there is not yet a database with audio files of different bees, for example of those with mite infestations. We could only train our prototype to distinguish between bees and ambient sounds. So the model still has to learn to recognize the specific sounds of bees infected with mites.
Nabil Basharat: But we have developed the one basic technique that can be installed on the hive to pick up sounds there and process them directly. It helped us a lot that we come from different backgrounds. Tobias Bartsch and Kevin are studying business informatics in their master's degree and took care of the software, the machine learning as well as the hardware architecture. In addition, Alexia Geoffrien was part of our team. She was an exchange student from France and incorporated the marketing perspective. I study industrial engineering and was mainly responsible for the design and manufacturing of the box that contains all the components and is attached to the beehive. Among other things, I used the 3D printer in the co-working space for this.
The new co-working space for start-up teams was an important part of the "Digital Innovation Lab" project format. How did you benefit from this creative space?
Basharat: What was especially great was that there is so much hardware there, such as the 3D printer. It also offers a lot of inspiration and potential for testing developments. And you can use the room flexibly and always have a place to meet at the university without having to search for a long time.
Kraus: Yes, the large-diaphragm microphone and the existing computer components for our microcontroller were also very helpful. Above all, however, we also used the room for exchanges with other groups.
Where do you go from here with your idea?
Kraus: We haven't had the chance to test it in practice yet, though. There are no plans to continue it after the end of the course so far, but our development is open source. That means that maybe there will be people interested in developing the prototype further and putting it to use.
The coworking space of the Management Transfer Lab
Since this year, the Management Transfer Lab (MTL) at the Faculty of Business Administration offers a space for joint, creative work to all those interested in founding a company from the university context as well as EXIST founding teams. The co-working space at Von-Melle-Park 5 offers a comprehensive range of hardware (including a 3D printer, 4K cameras, high-performance PCs and maker kits) and software (e.g. Qualtrics and Sawtooth). Founders will find space here for creative work, prototyping, validating their own business idea and exchanging ideas with other creatives. The space and hardware can be booked via the booking portal.
The coworking space as well as the project format "Digital Innovation Lab" offered by Prof. Dr. Jan Recker (Professorship for Information Systems and Digital Innovation) are funded by "Exist" - a program of the Federal Ministry for Economic Affairs and Climate Action (BMWK).