IEEE Spectrum AI→ original

Danish Scientists Created a Radar for Identifying Bee and Wasp Species

European scientists developed a millimeter-wave radar system that distinguishes bee species, wasps, and other pollinators by the micro-Doppler signatures of the

AI-processed from IEEE Spectrum AI; edited by Hamidun News
Danish Scientists Created a Radar for Identifying Bee and Wasp Species
Source: IEEE Spectrum AI. Collage: Hamidun News.
◐ Listen to article

Species identification of pollinators traditionally requires a dangerous and expensive method: insects are caught, killed, and examined in detail under a microscope. This is necessary for accurate identification, but harms populations. European researchers found a way to distinguish bees, wasps, and other pollinators without harm—using radar.

How the Radar Recognizes Insects

Scientists from Technological University of Denmark and Trinity College Dublin developed a system based on millimeter-wave radio. The key idea is simple: each insect species flaps its wings differently, and these movements create unique micro-Doppler signatures—specific patterns in radar reflection. It's like a fingerprint, but for wings.

Conventional radar has long been used only to track large swarms of migrating insects at high altitudes—for example, locusts or butterflies during migration. But the signal from a single small insect flying low over a flower is extremely weak. Adam Narbutovic, the research leader from Denmark's Technical University, explains: it was impossible to detect such a weak signal by simply looking at data from a single moment in time. The solution was found in signal integration: instead of analyzing a single moment in time, the system accumulates and processes data longer, extracting enough information for identification from it.

The specialists focused on how an insect's wing beats create micro-Doppler signatures—subtle changes in radar signal reflection caused by microscopic movements.

"When we look at raw signals, it's hard to catch all the subtle details. But with machine learning, we can distinguish them," — Adam Narbutovic.

Testing and Results

Scientists trained a machine learning model on five pollinator species: honeybees, bumblebees, and different types of wasps. The experiment was conducted on the Trinity College Dublin campus. Each insect was placed in a small plastic cylinder above a millimeter-wave antenna emitter, its radar signature was recorded, then released unharmed.

The model analyzed over 70 different characteristics of each insect's radar reflection, including wing-beat frequency, rate of motion change, and signal amplitude.

Results are impressive:

  • 85% accuracy in identifying a specific insect species
  • 96% accuracy in broader classification—distinguishing between bee and wasp groups
  • Analysis of 70+ characteristics of each radar reflection
  • Improved accuracy from 75% with 0.1 second observation to 84% with 1 second

Practical Applications

Researchers propose creating trap-like devices into which insects would naturally fly, where the system analyzes them in flight, and then they fly out unharmed. This opens up many practical applications.

Monitoring pollinator populations is critical for agriculture—bees pollinate about a third of the food we eat. The system can also track crop pests and detect invasive species before they spread.

The radio waves used in the system are completely safe—the power is well below any potentially harmful levels. This is starkly different from traditional traps, which often use cyanides or other toxic substances.

What This Means

The shift from killing insects to non-invasive monitoring is a huge step for entomology and biodiversity conservation. The scientists' next goal is to develop a portable version for field use and compile a global database of radar signatures of all known pollinators.

Such a database would allow instant identification of an insect by its flight pattern. By adding environmental data, it would be possible to track not just species composition, but also behavioral changes—for example, anomalous patterns in wing-beat frequency that signal population stress or disease.

ZK
Hamidun News
AI news without noise. Daily editorial selection from 400+ sources. A product by Zhemal Khamidun, Head of AI at Alpina Digital.

Want to stop reading about AI and start using it?

AI News is a curated feed of AI/tech news. Hamidun Academy teaches you to use AI systematically in your work.

What do you think?
Loading comments…