Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • ADVERTISEMENT FEATURE Advertiser retains sole responsibility for the content of this article

Capturing snapshots of biodiversity

Anomaly processing and detection of animals using a prototype wildlife surveillance system.© Bernd Radig

A wildlife surveillance system that can accurately detect and identify animal species using smart algorithms has been developed by researchers in Germany. The low-power, self-sufficient system, reported in Pattern Recognition and Image Analysis, offers an efficient way to measure changes in biodiversity with minimal human intervention.

In a changing climate, it has never been more important to keep track of biodiversity. But monitoring these ecosystem changes is difficult, with experts relying on human observations and specialized equipment to measure species abundance and diversity.

While camera traps have become an essential tool for capturing wildlife and their movements, they generate enormous amounts of data. “The analysis of the millions of images collected by camera traps is time-consuming, if not impossible, for specialists,” says lead author, Bernd Radig, at the Technical University of Munich, in Germany.

Radig and colleagues designed a prototype wildlife surveillance system for the German government’s Automated Multisensor Stations for Monitoring of BioDiversity (AMMOD) project, which is aiming to install a network of automated biodiversity ‘weather stations’ in national parks across the country.

The camera system includes two lenses to capture three-dimensional images of animals in daylight and night, a motion detector, and a processor that uses depth information to separate wildlife from background features. The platform’s computer pre-processes the images and sends them to a server when enough bandwidth and power are available. Once there, deep-learning algorithms identify the animal species in the images, saving researchers from ploughing through thousands of camera-trap photos. The team trained these species-detecting algorithms using collections of labelled images in museum and citizen science databases.

“We engineered these platforms to be very low-energy but reliable, so that maintenance is required no more than every six months,” says Radig.

In addition to the camera system, the surveillance platform also includes a scanner that takes high-resolution images of moths that land on an illuminated screen. Tests of the system revealed that it detected moths with around 94% accuracy and correctly identified species 89% of the time.

To cut down the volume of transmitted data, the platform’s local computer only sends images containing moths to the server. The next step for Radig and his team is to find a way to process data on-site, while saving energy. “It’s difficult to find a balance between power consumption and transferring enough data,” says Radig.

Radig and his colleagues will soon begin trialling their wildlife surveillance system at stations in national parks across Germany. “We have hundreds of stations that deliver data about weather conditions, but we also need stations for monitoring biodiversity,” says Radig.

This collection of research highlights is produced by the Partnership & Custom Media unit of Nature Research for Pleiades Publishing. The advertiser retains responsibility for content.

Read the original research article for free here.


  1. Radig, B., Bodesheim, P., Korsch, D. et al. Automated visual large scale monitoring of faunal biodiversity. Pattern Recognit. Image Anal. 31, 477–488 (2021).

    Article  PubMed  Google Scholar 

Download references

Related Articles


Quick links